Magnetism and Structure in Functional Materials (Springer Series in Materials Science)

Citation preview

Springer Series in

materials science

79

Springer Series in

materials science Editors: R. Hull

R. M. Osgood, Jr.

J. Parisi

H. Warlimont

The Springer Series in Materials Science covers the complete spectrum of materials physics, including fundamental principles, physical properties, materials theory and design. Recognizing the increasing importance of materials science in future device technologies, the book titles in this series ref lect the state-of-the-art in understanding and controlling the structure and properties of all important classes of materials. 71 Dissipative Phenomena in Condensed Matter Some Applications By S. Dattagupta and S. Puri

79 Magnetism and Structure in Functional Materials Editors: A. Planes, L. Ma˜nosa, and A. Saxena

72 Predictive Simulation of Semiconductor Processing Status and Challenges Editors: J. Dabrowski and E.R. Weber

80 Ion Implantation and Synthesis of Materials By M. Nastasi and J.W. Mayer

73 SiC Power Materials Devices and Applications Editor: Z.C. Feng 74 Plastic Deformation in Nanocrystalline Materials By M.Yu. Gutkin and I.A. Ovid’ko 75 Wafer Bonding Applications and Technology Editors: M. Alexe and U. G¨osele 76 Spirally Anisotropic Composites By G.E. Freger, V.N. Kestelman, and D.G. Freger 77 Impurities Confined in Quantum Structures By P.O. Holtz and Q.X. Zhao 78 Macromolecular Nanostructured Materials Editors: N. Ueyama and A. Harada

81 Metallopolymer Nanocomposites By A.D. Pomogailo and V.N. Kestelman 82 Plastics for Corrosion Inhibition By V.A. Goldade, L.S. Pinchuk, A.V. Makarevich and V.N. Kestelman 83 Spectroscopic Properties of Rare Earths in Optical Materials Editors: G. Liu and B. Jacquier 84 Hartree–Fock–Slater Method for Materials Science The DV–X Alpha Method for Design and Characterization of Materials Editors: H. Adachi, T. Mukoyama, and J. Kawai 85 Lifetime Spectroscopy A Method of Defect Characterization in Silicon for Photovoltaic Applications By S. Rein 86 Wide-Gap Chalcopyrites Editors: S. Siebentritt and U. Rau

Volumes 20–70 are listed at the end of the book.

A. Planes (Eds.)

L. Ma˜nosa

A. Saxena

Magnetism and Structure in Functional Materials With 129 Figures (8 in color)

123

Professor Antoni Planes ´ Manosa ˜ Professor Lluıs Universitat de Barcelona, Facultat de F´ısica Departament d’Estructura i Constituents de la Mat`eria Diagonal 647, 08028 Barcelona, Catalonia, Spain E-mail: [email protected], [email protected]

Dr. Avadh Saxena Los Alamos National Lab, Theoretical Division, T-11, MS B262 Los Alamos, NM 87545, USA E-mail: [email protected]

Series Editors:

Professor Robert Hull

Professor Jürgen Parisi

University of Virginia Dept. of Materials Science and Engineering Thornton Hall Charlottesville, VA 22903-2442, USA

Universit¨at Oldenburg, Fachbereich Physik Abt. Energie- und Halbleiterforschung Carl-von-Ossietzky-Strasse 9–11 26129 Oldenburg, Germany

Professor R. M. Osgood, Jr.

Professor Hans Warlimont

Microelectronics Science Laboratory Department of Electrical Engineering Columbia University Seeley W. Mudd Building New York, NY 10027, USA

Institut f¨ur Festk¨orperund Werkstofforschung, Helmholtzstrasse 20 01069 Dresden, Germany

ISSN 0933-033X ISBN-10 3-540-23672-4 Springer Berlin Heidelberg New York ISBN-13 978-3-540-23672-6 Springer Berlin Heidelberg New York Library of Congress Control Number: 2004117013 This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned, specif ically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microf ilm or in any other way, and storage in data banks. Duplication of this publication or parts thereof is permitted only under the provisions of the German Copyright Law of September 9, 1965, in its current version, and permission for use must always be obtained from Springer-Verlag. Violations are liable to prosecution under the German Copyright Law. Springer is a part of Springer Science+Business Media. springer.com © Springer-Verlag Berlin Heidelberg 2005 Printed in Germany The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply, even in the absence of a specif ic statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. Typesetting by the Authors and SPI Publisher Services using a Springer TEX macro package Cover concept: eStudio Calamar Steinen Cover production: design & production GmbH, Heidelberg Printed on acid-free paper

SPIN: 10948583

57/3100/ SPI

543210

In Memorium James Arthur Krumhansl

(August 2, 1919–May 6, 2004)

Preface

This book is based on the emerging notion that a thorough and fundamental understanding of the interplay of magnetism and structure is needed if a wide variety of functional materials are to be efficiently used in applications. The intent is to foster interaction between researchers in three distinct but related topics: (i) magnetoelastic materials such as magnetic martensites and magnetic shape memory alloys, (ii) magnetocaloric effects related to magnetostructural transitions, and (iii) colossal magnetoresistance (CMR) manganites and related perovskites. The goal is to identify common underlying principles in these three classes of materials that are relevant to optimize various functionalities. The interplay between structure and magnetic properties is sensitive to disorder and manifests itself at multiple length scales in a cross-correlated and hierarchical fashion. Functional materials are becoming a major theme of the XXI century research. The emergence of apparently different magnetic/structural phenomena in disparate classes of materials (mentioned above) clearly points to a need to integrate common concepts in order to achieve a broader understanding of the interplay between magnetism and structure in this general class of functional materials. This book represents the first steps toward this goal and we hope it reflects this at least in spirit. The idea for this book was born at the workshop Interplay of Magnetism and Structure in Functional Materials held at the Benasque Center for Science in the Pyrenees mountains during February, 9–13, 2004. The topics covered here are interdisciplinary in nature and the chapters are written by expert researchers, who are from physics, materials science, applied mathematics and engineering backgrounds. Therefore, the book is addressed to both the experts and researchers getting into the field of magnetic functional materials including graduate students. The eleven chapters are arranged as follows. The first two chapters (respectively by Shenoy et al. and Cast´ an et al.) focus on the general aspects of phase transitions in multiferroic materials such as the role of elastic compatibility and long-range interactions, coupling between strain and other impor-

VIII

Preface

tant physical variables, e.g. magnetization, polarization, orbital ordering, etc., role of disorder in pretransitional phenomena and transition kinetics, local structure probes and phonon anomalies. The next two chapters by Morellon and Ibarra, and by Egami, cover a broad variety of magnetoresistive, magnetostrictive and superconducting materials where the authors discuss the role played by the interplay of spin, charge, orbital and lattice degrees of freedom on their properties. The chapters by Shapiro, Venkateswaran and De Graef, and Brown et al., review and provide exciting new results concerning the anomalous phonon behavior, structural and magnetic precursors and techniques for imaging them. James and Zhang provide an applied mathematical perspective on the problem of martensites as well as the conditions for the coexistence of ferroelectricity and magnetism. The century old (but still not fully understood) problem of the Invar as well as anti-Invar effects are reviewed by Wasserman and Acet. Finally, Pecharsky and Gschneider Jr. and Casanova et al. elaborate on the magnetocaloric effects and emerging new materials and applications. The last chapter by Prof. P.-A. Lindg˚ ard nicely summarizes the workshop held in Benasque and provides perspectives for the future. We would like to dedicate this book to the memory of Prof. James A. Krumhansl who passed away while writing the Introductory chapter. This unfinished chapter with his thoughts on the complex functional materials is also included in the beginning of this book. Over six decades Prof. Krumhansl inspired many generations of scientists in the multidisciplinary field of complex materials and brought the concept of (elastic) solitons in materials science as simply being domain walls or twin boundaries. He also popularized the physicist’s notion of phonons among metallurgists which has undoubtedly helped the interpretation of many experimental observations. Finally, we gratefully acknowledge financial support from MCyT (Spain), AGAUR (Catalonia), University of Barcelona, Los Alamos National Laboratory, the U.S. Department of Energy and Benasque Center for Science. We specially thank Prof. Pere Pascual (Director of Benasque Center) for gracefully providing the conference facilities at Benasque.

Barcelona and Los Alamos, October 2004

Antoni Planes Llu´ıs Ma˜ nosa Avadh Saxena

Contents

1 Complex Functional Materials J.A. Krumhansl . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1

2 Spin, Charge, and Lattice Coupling in Multiferroic Materials S.R. Shenoy, T. Lookman, and A. Saxena . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2 Order Parameters and Multiferroics . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3 Strain Tensor and Elastic Compatibility . . . . . . . . . . . . . . . . . . . . . . . 2.3.1 Levels of Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3.2 Definition of Unit-Cell Strain Tensor . . . . . . . . . . . . . . . . . . . . 2.3.3 Strain Free Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3.4 St Venant Compatibility Constraints . . . . . . . . . . . . . . . . . . . . 2.3.5 Compatibility Potentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3.6 Other Approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.4 Inhomogeneities in Ferroelastics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.5 Inhomogeneities in Multiferroic Oxides . . . . . . . . . . . . . . . . . . . . . . . . 2.6 Charge and Spin as Local Stresses and Transition Temperatures . . 2.6.1 Coupling of Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.6.2 Simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.7 Summary and Further Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3 3 5 6 6 7 8 9 10 12 13 16 18 18 19 21 22

3 Disorder in Magnetic and Structural Transitions: Pretransitional Phenomena and Kinetics T. Cast´ an, E. Vives, L. Ma˜ nosa, A. Planes, and A. Saxena . . . . . . . . . . 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2 Disorder Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.1 Pretransitional Phenomena . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.2 Premonitory Behaviour: Temperature Softening . . . . . . . . . . 3.2.3 Structural Precursors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

27 27 28 29 29 31

X

Contents

3.2.4 Coupling to Magnetism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.5 Ferroic Precursors: Magnetic and Others . . . . . . . . . . . . . . . . 3.3 Transition Kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.1 Athermal Transitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.2 Modelling Athermal Transitions . . . . . . . . . . . . . . . . . . . . . . . . 3.3.3 Avalanche Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.4 Modelling Avalanches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Huge Magnetoresistance in Association with Strong Magnetoelastic Effects L. Morellon and M.R. Ibarra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2 Magnetic–Crystallographic Transformations . . . . . . . . . . . . . . . . . . . . 4.2.1 Gd5 (Six Ge1−x )4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.2 MnAs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3 Moment Instabilities and Spin Fluctuations . . . . . . . . . . . . . . . . . . . . 4.3.1 FeRh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.2 Hf1−x Tax Fe2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.3 La(Fex Al1−x )13 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4 Metal–Insulator Transitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.1 Colossal Magnetoresistance (CMR) and Huge Volume Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.2 Magnetostriction and Magnetoresistance in the Paramagnetic Phase . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.3 Magnetostriction in Inhomogeneous Electronic Systems: Static Phase Segregation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.5 Charge/Orbital Instabilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 Interplay of Spin, Charge, and Lattice in CMR Manganites and HTSC Cuprates T. Egami . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2 Stability of Polaronic Phase in the CMR Manganites . . . . . . . . . . . . 5.2.1 Colossal Magnetoresistivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2.2 Pulsed Neutron PDF Method . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2.3 Polaron Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3 Stability of Spin–Charge Stripes in the Cuprates . . . . . . . . . . . . . . . . 5.4 Electron–Phonon Coupling and Mechanism of Superconductivity in the Cuprates . . . . . . . . . . . . . . . . . . . . . . . . . . 5.4.1 Local Lattice Distortion and Inhomogeneous Electronic States . . . . . . . . . . . . . . . . . . .

33 35 37 37 40 40 43 45 46

49 49 50 50 51 53 53 55 56 57 59 61 63 65 70 71

75 75 76 76 77 79 83 84 84

Contents

5.4.2 5.4.3

Electron–Phonon Coupling in the Cuprates . . . . . . . . . . . . . . Vibronic Mechanism of High Temperature Superconductivity . . . . . . . . . . . . . . . . . 5.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

XI

84 87 89 90

6 Neutron Scattering Studies of Anomalous Phonon Behavior in Functional Materials S.M. Shapiro . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93 6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93 6.2 Neutron Scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94 6.3 Phonon Anomalies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95 6.3.1 Kohn Anomalies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96 6.3.2 Soft Modes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97 6.4 Phonon Anomalies in the Manganites . . . . . . . . . . . . . . . . . . . . . . . . . . 99 6.5 Phonon Anomalies in High Temperature Superconductors . . . . . . . . 100 6.6 Ferromagnetic Shape Memory Alloys . . . . . . . . . . . . . . . . . . . . . . . . . . 102 6.6.1 Iron-Based Alloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103 6.6.2 Heusler-Based Alloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104 6.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111 7 The Structures and Transformation Mechanism in the Ferromagnetic Shape Memory Alloy Ni2 MnGa P.J. Brown, T. Kanomata, M. Matsumoto, K.-U. Neumann, and K.R.A. Ziebeck . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113 7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113 7.2 The Crystal Structure of the Cubic Austenite Phase . . . . . . . . . . . . . 113 7.3 Bulk Magnetic Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116 7.4 Spin Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118 7.5 Paramagnetic Response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118 7.6 Inelastic Neutron Scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119 7.7 Neutron Diffraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121 7.8 Pre-Martensitic Phase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122 7.8.1 Field Dependence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125 7.9 The Martensitic Phase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126 7.10 Structural and Magnetic Phase Diagram . . . . . . . . . . . . . . . . . . . . . . . 127 7.11 Mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129 7.12 Martensitic Twinning in Ni2 MnGa . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130 7.13 Non-Stoichiometric Samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133 7.14 Electron Concentration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135 7.15 Polarised Neutron Scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136 7.16 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138

XII

Contents

8 Imaging Techniques in Magnetoelastic Materials S.P. Venkateswaran and M. De Graef . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141 8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141 8.2 Lorentz Image Formation Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141 8.2.1 Classical Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141 8.2.2 Quantum Mechanical Description . . . . . . . . . . . . . . . . . . . . . . . 143 8.2.3 Phase Reconstruction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145 8.3 Applications of LTEM to Ferromagnetic Shape Memory Alloys . . . 148 8.3.1 Ni2 MnGa, Austenitic State . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149 8.3.2 Ni2 MnGa, Martensitic State . . . . . . . . . . . . . . . . . . . . . . . . . . . 152 8.3.3 Co2 NiGa, Austenitic State . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153 8.3.4 Co2 NiGa, Martensitic State . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154 8.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156 9 A Way to Search for Multiferroic Materials with “Unlikely” Combinations of Physical Properties R.D. James and Z. Zhang . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159 9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159 9.2 Single Phase Multiferroics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160 9.3 Basic Idea . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160 9.4 Lattice Parameter Sensitivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161 9.5 What Makes Big First Order Phase Transformations Reversible? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162 9.6 Specific Relationships Among Lattice Parameters for a High Degree of Reversibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169 9.6.1 Cubic to Orthorhombic Transformations . . . . . . . . . . . . . . . . . 170 9.6.2 Cubic to Monoclinic Transformations . . . . . . . . . . . . . . . . . . . 170 9.6.3 Relationships for Martensite/Martensite Transitions . . . . . . 171 9.7 Tuning Lattice Parameters to Satisfy Two of the Proposed Conditions in the NiTiCuPd System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171 9.8 Further Comparisons with Experiment . . . . . . . . . . . . . . . . . . . . . . . . . 172 9.9 Summary and Outlook: A General Method for Seeking New Classes of Functional Materials . . . . . . . . . . . . . . . . 174 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174 10 Invar and Anti-Invar: Magnetovolume Effects in Fe-Based Alloys Revisited E.F. Wassermann and M. Acet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177 10.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177 10.2 Invar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177 10.3 From Invar to Anti-Invar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181 10.4 Allotropy of Pure Fe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184 10.5 Ground State Properties of Invar and Anti-Invar . . . . . . . . . . . . . . . . 187

Contents

XIII

10.6 Pressure Experiments: Evidence for High Spin to Low Spin State Transitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189 10.7 HS–LS Transitions in a Microscopic Picture . . . . . . . . . . . . . . . . . . . . 192 10.8 Questions and Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196 11 Magnetocaloric Effect Associated with Magnetostructural Transitions V.K. Pecharsky and K.A. Gschneidner, Jr. . . . . . . . . . . . . . . . . . . . . . . . . . . 199 11.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199 11.2 Magnetic Cooling or Why Having a Strong Magnetocaloric Effect in a Weak Magnetic Field Makes a Difference? . . . . . . . . . . . . . . . . . . 199 11.3 Gd5 (Si4−x Gex ) System and the Giant Magnetocaloric Effect . . . . . . 202 11.4 Altering Crystal Structures with a Magnetic Field . . . . . . . . . . . . . . . 206 11.5 To What Extent a Structural Change Enhances the Giant Magnetocaloric Effect? . . . . . . . . . . . . . . . . . . . . 214 11.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220 12 Entropy Change and Magnetocaloric Effect in Magnetostructural Transformations F. Casanova, X. Batlle, A. Labarta, J. Marcos, E. Vives, L. Ma˜ nosa, and A. Planes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223 12.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223 12.2 Multiscale Origin of the MCE in Ni–Mn–Ga Alloys . . . . . . . . . . . . . . 225 12.3 Direct Determination of the Entropy Change at a First-Order Transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229 12.4 Magnetostructural Transformation in Gd–Si–Ge Alloys . . . . . . . . . . 230 12.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235 13 Functional Magneto-Structural Materials: Summary and Perspectives P.-A. Lindg˚ ard . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243 Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247

List of Contributors

M. Acet Experimentalphysik Universit¨at Duisburg–Essen D-47048 Duisburg, Germany [email protected]

X. Batlle Departament de F´ısica Fonamental Universitat de Barcelona Diagonal 647 08028 Barcelona, Spain [email protected]

P.J. Brown Institut Laue Langevin 38042 Grenoble, France and Department of Physics Loughborough University Loughborough LE11 3TU, UK [email protected]

F. Casanova Departament de F´ısica Fonamental Universitat de Barcelona Diagonal 647 08028 Barcelona, Spain [email protected]

T. Cast´ an Departament d’Estructura i Constituents de la Mat`eria Facultat de F´ısica Universitat de Barcelona Diagonal 647 08028 Barcelona, Spain [email protected]

M. De Graef Department of Materials Science and Engineering Carnegie Mellon University 5000 Forbes Avenue Pittsburgh, PA 15213-3890, USA [email protected]

T. Egami University of Tennessee Departments of Materials Science, Engineering, Physics, and Astronomy Knoxville, TN 37996, USA Oak Ridge National Laboratory Oak Ridge, TN 37831, USA [email protected]

XVI

List of Contributors

K.A. Gschneidner, Jr. Ames Laboratory US Department of Energy Department of Materials Science and Engineering Iowa State University Ames, IA 50011-3020, USA [email protected] M.R. Ibarra Departamento de F´ısica de la Materia Condensada and Instituto de Ciencia de Materiales de Arag´on Universidad de Zaragoza and Consejo Superior de Investigaciones Cient´ıficas 50009 Zaragoza, Spain [email protected] R.D. James Department of Aerospace Engineering and Mechanics University of Minnesota Minneapolis, MN 55455, USA [email protected] T. Kanomata Faculty of Engineering Tohoku Gakuin University Tagajo 985, Japan [email protected]. gakuin.ac.jp A. Labarta Departament de F´ısica Fonamental Universitat de Barcelona Diagonal 647 08028 Barcelona, Spain [email protected] P.-A. Lindg˚ ard Materials Research Department Risø National Laboratory 4000 Roskilde, Denmark [email protected]

T. Lookman Theoretical Division Los Alamos National Lab Los Alamos, NM 87545, USA [email protected] J. Marcos Departament d’Estructura i Constituents de la Mat`eria Universitat de Barcelona Diagonal 647 08028 Barcelona, Spain [email protected] L. Ma˜ nosa Departament d’Estructura i Constituents de la Mat`eria Facultat de F´ısica Universitat de Barcelona Diagonal 647 08028 Barcelona, Spain [email protected] M. Matsumoto IMRAM, Tohoku University Sendai 980-8577, Japan [email protected] L. Morellon Departamento de F´ısica de la Materia Condensada and Instituto de Ciencia de Materiales de Arag´ on Universidad de Zaragoza and Consejo Superior de Investigaciones Cient´ıficas, 50009 Zaragoza, Spain [email protected] K.-U. Neumann Department of Physics Loughborough University Loughborough LE11 3TU, UK [email protected]

List of Contributors

V.K. Pecharsky Ames Laboratory US Department of Energy Department of Materials Science and Engineering Iowa State University Ames, IA 50011-3020, USA [email protected] A. Planes Departament d’Estructura i Constituents de la Mat`eria Facultat de F´ısica Universitat de Barcelona Diagonal 647 08028 Barcelona, Spain [email protected] A. Saxena Theoretical Division Los Alamos National Laboratory Los Alamos, NM 87545, USA [email protected] S.M. Shapiro Department of Physics Brookhaven National Laboratory Upton, NY 11973, USA [email protected] S.R. Shenoy International Centre for Theoretical Physics Trieste, Italy [email protected]

XVII

S.P. Venkateswaran Department of Materials Science and Engineering Carnegie Mellon University 5000 Forbes Avenue Pittsburgh, PA 15213-3890, USA [email protected] E. Vives Departament d’Estructura i Constituents de la Mat`eria Facultat de F´ısica Universitat de Barcelona Diagonal 647 08028 Barcelona, Spain [email protected] E.F. Wassermann Experimentalphysik Universit¨at Duisburg–Essen D-47048 Duisburg, Germany [email protected] Z. Zhang Department of Aerospace Engineering and Mechanics University of Minnesota Minneapolis, MN 55455, USA [email protected] K.R.A. Ziebeck Department of Physics Loughborough University Loughborough LE11 3TU, UK [email protected]

1 Complex Functional Materials J.A. Krumhansl [Based on notes provided by Prof. J.A. Krumhansl to the editors.]

To begin with, I want to say that the idea that complex functional/adaptive materials are a new, recent discovery displays a shallow knowledge of metallurgy, last century’s condensed matter physics, and electrical engineering. But to get on to the point of the title: Some ten years ago when many complex . . . materials were being widely publicized for new funding, Dick Slansky, then Director of the Theoretical Division at Los Alamos National Laboratory, was complaining to me: “What good does ‘complex’, defined as non-simple, do for us scientifically? What kind of a guiding principle for science is that?” To our great surprise we found that the main definition of “complex” has a well-defined meaning and is very useful to generalize to materials. Simple, we found the answer in most dictionaries! To me it is a most conceptually useful definition, and I hope you find it so too: Complex (def. 1): “Composed of interconnected or interwoven parts”. . . This is the main meaning. And it clearly describes many early Complex Functional Materials, such as a ferromagnet. That early work identified multiple domains which moved at the mesoscopic scale in response to an applied field by motion of magnetic domain walls. The effect is to change the average macroscopic magnetization. So this definition describes an array of interwoven microscopic phases which move, respectively, to change a macroscopic condition. Not surprisingly the same ideas apply to ferroelectrics and ferroelastics (martensites) and shape memory materials. The materials discussed in this book now can also be thought of in this context. I offer some miscellaneous remarks on modern materials science. An important reference is “Renormalization Group Theory: Its basis and formulation in statistical physics”, Michael E. Fisher, Reviews of Modern Physics 70, p. 653 (1998). Read particularly pp. 653–655 and consider the following questions (1) Is the Renormalization Group (RG) a fundamental principle or a calculation method? (2) What fraction of all phase transitions are second order? Is Renormalization Group needed generally in Materials Science? (3) Is there really such a thing as First Principles calculations (at finite temperature)? What

2

J.A. Krumhansl

is an effective Hamiltonian in modern condensed matter physics? I provide a few hints and partial answers. A vast majority (more than 90%) of all phase transitions in nature are of first order. Renormalization Group is essentially a calculational procedure. However, the Landau theory is of fundamental importance in materials science and so forth. The key word for modeling is multiscale. Physically realistic models can start at the mesoscopic scale where the notion of an order parameter and the corresponding Ginzburg–Landau free energy (for phase transitions) are essential. The coefficients of this free energy must be determined from experimental data such as structural, phonon dispersion, and thermodynamic response functions. Alternatively, these coefficients can be determined from microscopic calculations, e.g., electronic structure calculations and molecular dynamics simulations. The single crystal Ginzburg–Landau models can be extended to polycrystals and then suitably coarse grained to serve as input for continuum level constitutive response models. Over the past half century I have had spirited discussions and shared many ideas with most of the authors in this book, and my own understanding of complex materials and multiferroics, using martensites as a prototypical example, has evolved. Understanding is always a relative word and I believe that the younger generation of scientists will take it to the next level by exploiting the modern tools of computational science, high resolution microscopies, and a suite of characterization techniques. The path to new ideas and applications in the field of functional materials is multidisciplinary – a synergy between metallurgists, materials scientists, physicists, applied mathematicians, engineers, and even biologists. We have to learn a lot from biology in terms of how evolution has optimized biological functionality.

2 Spin, Charge, and Lattice Coupling in Multiferroic Materials S.R. Shenoy, T. Lookman, and A. Saxena

2.1 Introduction Functional materials are those with technologically useful properties that can be sensitively controlled by external parameters. The need to understand and to use functionalities of shape memory alloys, relaxor ferroelectrics, colossal magnetoresistance manganites, superconducting cuprates and other materials, is one motivation for their intense current study [1–11]. Of course, applied stress, magnetic field or voltage will, respectively, induce direct-variable changes in strain, magnetization and electric polarization. More interestingly, in multifunctional materials with mutually coupled variables, there is also a cross-variable response, so a given field could control two or more variables, or a variable can be switched by two or more external fields. Thus in magnetic shape memory alloys, with coupled strain and magnetization, the external shape can be controlled both by pressure and by magnetic field; and stress changes magnetization, while magnetic field affects structure [12]. In colossal magnetoresistance manganites [3], the conductance (that is related to magnetization), rises by many orders of magnitude in response to uniform magnetic fields as well as to hydrostatic pressure; while voltage induces changes in local structure and conductance [13, 14]. In ferroelectric manganites, polarization can be switched by a magnetic field [15]. In cuprates [7] the superconductivity is suppressed as usual by a magnetic field, but is also controlled by stress [16]; shape memory-like effects can occur [17]; and surprisingly (for these doped antiferromagnets), an external uniform magnetic field can rotate elastic domain walls [18]. Thus a variety of materials show cross-variable response in their macroscopic multifunctionalities, implying strong coupling of magnetic spin, electronic charge and lattice strain. Finescale local probes have revealed something surprising, however. Many transition metal oxides spontaneously form nonuniform complex patterns of charge, spin and strain variables at scales of a nanometer and upwards [4–6, 19–26]. The multivariable patterning includes structural stripes,

4

S.R. Shenoy et al.

magnetic droplets and conducting filaments, and there has been much experimental and theoretical interest in this “nanoscale phase separation” or “intrinsic heterogeneity” [4–6]. There is also a remarkable cross-scale response: a uniform or macroscale magnetic field can induce mesoscale structural changes, or nanoscale conductance variations [18, 19]. An understanding of these multiscale, multivariable textures is crucial to tailoring multifunctionalities on the macroscale [10]. Clearly, a uniform order parameter (OP) Landau model is insufficient. The spatially varying patterns of charge, magnetization, and strain must emerge as the locally stable minimum of some multivariable, nonlinear, free energy functional. Spatial variation as a preferred state is not new, of course. Purely ferroelastic materials with a strain tensor order parameter, such as shape memory alloys like FePd, show mesoscale “twins” or bands of coexisting unit-cell variants, below the structural transition [27]; or woolen-fabric-like “tweed” of oriented criss-cross striations on a finer nanoscale, above the transition [28]. In martensites, interfaces can also be dynamic, with acoustic emission on quenching, from domains advancing at sound speeds [29]. Superconducting cuprates are also doped ferroelastics [30], and show both twinning and tweed, as emphasized by Krumhansl [7]. Magnetoresistant metallic glasses exhibit mesoscale tweed [31]. Relaxor ferroelectrics/magnetoelastics also show twins and tweed in electric polarization/magnetization [12, 32]. The similarity and ubiquity of textures in very different materials prompts a search for a common link. Competition between short- and long-range forces, and/or sign-varying frustration effects are known to generically favour patterning [33]. Since all variables live on a discrete, d-dimensional crystal lattice, it follows that local deformation or strain of the lattice both affects spin and charge, and in turn is influenced by them. It turns out that in ferroelastics, generic lattice-integrity constraints on the 12 d(d + 1) components of the symmetric strain tensor [34] can induce effective long-range interaction potentials, with sign-varying angular variation, and power-law decay. Working in a strain representation, rather than in the more conventional displacement representation, reveals these hidden, texture-inducing correlations. The origins of intrinsic inhomogeneities could lie in the ubiquitous strain tensor: its power-law interaction, its nonlinearity, and its coupling to other variables. We focus on this emerging viewpoint [35–42], without giving a proper review of other approaches, or any more than illustrative references. In Sect. 2.2 we consider coupled “multiferroic” order parameters. Section 2.3 discusses the St Venant compatibility constraint that induces the effective long-range potential. Intrinsic inhomogeneities in proper ferroelastics and multiferroic oxides, are discussed in Sects. 2.4 and 2.5, respectively. Section 2.6 outlines the modelling of oxides, with charge and spin acting as local stresses and temperatures. A summary, with speculations on further work, is given in Sect. 2.7.

2 Spin, Charge, and Lattice Coupling in Multiferroic Materials

5

2.2 Order Parameters and Multiferroics The order parameter is a basic concept in phase transitions [43]: a variable whose thermal average becomes nonzero below a transition temperature or pressure either continuously (second-order) or discontinuously (first-order). The free energy lowering can involve a global dominance of a uniform state: only one out of NV competing, lower symmetry, degenerate minima or “variants” is finally present. Nonlinearity of the free energy functional in such second order transitions is manifest close to the transition, where renormalizations occur due to nonuniform critical fluctuations of competing variants. Structural phase transitions involve a change in the discrete symmetry of the unit cell, with lattice points shifting on cooling (or application of stress) to new lower-symmetry positions. Ferroelastic displacive structural transitions [1] are those for which the symmetry of the low-temperature unit cell is a subgroup of the high-temperature unit cell symmetry. The order parameters (OP) for first-order ferroelastic transitions are NOP ≤ d independent components of the strain tensor, that can exhibit strain hysteresis under applied stress. The free energy lowering can involve a global coexistence of nonuniform states: all the NV lower symmetry variants are finally present. Nonlinearity of the free energy functional in such first-order transitions is manifest for temperatures all the way below, and for a range above, the transition. Structural static domain walls such as in twins or tweed, and dynamic excitations, are elastic solitons [27,28,44], and the nonuniformity is a signature of nonlinearity. Domains and hysteretic switching occur, of course, in several types of materials – indeed the very name “ferroelastic” and “ferroelectric” derive from the analogy with ferromagnets: they all have such behaviour in common (but not the element of Fe, ironically!) These three materials classes can be grouped under the common rubric of “ferroics” as they all “undergo nondisruptive phase transitions that alter their point group symmetry” with strain, electric polarization, or magnetization as the respective order parameters [9]. Materials with two or more ferroic OPs, and a sequence of nested phase transitions, are multiferroics, that can have mutually coupled nonlinearities. A primary order parameter at its transition temperature (or composition), can drive the other, secondary OPs; with roles exchanged at the other, nested transitions. For example cuprates have a ferroelastic structural transition [30], with strain as the primary OP, that can be followed on cooling by a N´eel transition, with staggered magnetization as the primary OP [7]. In manganites depending on the doping, there is a complex phase diagram, with ferromagnetic, chargeorder, ferroelastic and Jahn-Teller transitions [3, 6]. As illustrated in later sections, cuprates and manganites could be regarded as examples of complex multiferroics in the sense of Krumhansl, who speaks in this volume of complex materials as composed of interconnected parts, with interwoven microscopic phases moving relatively, to change a macroscopic condition. This picture of annealed-variable texturing

6

S.R. Shenoy et al.

is complementary to quenched-disorder effects, certainly also present [6]. The multiferroic conceptual umbrella, and an approach that studies cuprates in the context of manganites and other oxides [23], might have pleased Krumhansl, who also remarked that “by and large, the physics literature on high Tc seems to be unfortunately disjoint from much earlier work on displacive phase transitions, or other areas of materials science, notably ferroelectricity and metallurgy” [7].

2.3 Strain Tensor and Elastic Compatibility In this section we focus on “proper” ferroelastics that have only a strain tensor order parameter [1] and later consider couplings to other variables, in “improper” ferroelastics. Suppose we only consider those deformations of the crystal lattice that do not create dislocations and vacancies. Then the central point is that this seemingly innocuous “no-defect” constraint, of lattice integrity maintained at every instant, implies the existence of anisotropic, power-law (APL) correlations between order-parameter strains. 2.3.1 Levels of Description The unit cell sets the minimum scale for strain variations. More generally, it is useful to define levels of description for probe distances R, as in Fig. 2.1. Level 0 or L0 ∼ R  a0 (lattice constant) is the macroscopic regime of engineering strains, and of system size and shape. Level 1 is the regime of unit-cell strains, that can have meso- and nanoscale variations within the large range L0 > R > a0 . The intracell regime of Level 2 involves the bond angles and lengths, a0 > R ∼
0 , of the atomic basis whose deformation

Fig. 2.1. Levels of description. See text, Sect. 2.3.1

2 Spin, Charge, and Lattice Coupling in Multiferroic Materials

7

may be termed as the “microstrain”, e.g. Jahn–Teller-like distortions of perovskite octahedra [21]. Finally, Level 3 is the ˚ Angstrom-scale regime of the electron occupancies/spin states of individual-atom wavefunctions on scales
0 > R ∼ r0 , e.g. d-wave orbitals or core spins. Of course, the levels are coupled. The Level 1/Level 2 interface is where the elastic physics meets the structural chemistry. Jahn–Teller Hamiltonians [3] link Level 3 orbital occupancies to Level 2 octahedral “Q2 , Q3 ” stretchings; these in turn represent Level 1 unit-cell strain distortions. The material specifics of the multiferroics are in the detailed nature of the charge/polarization/spin variables; their symmetry-dictated couplings to each other and to strains; and the values and signs of the coupling constants as well as energy scales. But for materials with strong elastic anisotropy (from directional Level 2 bonding, or basis atoms with widely different sizes), the common key is the Level 1 strain. Its APL elastic correlations, nonlinearities, and couplings to other variables, could possibly induce domain-wall patterned states, as local minima of a multivalley free energy landscape. The power-law correlations are in a certain sense, scale-free, and could link widely separated variables at finer levels of description. 2.3.2 Definition of Unit-Cell Strain Tensor Engineering strain is a Level 0 description of distortions of a medium, with derivatives of a continuous displacement u(r
), labelled by a continuum position r
[43,45]. On the other hand, unit-cell strain in a crystal lattice with lattice points r i can be defined as a Level 1 continuous variable, that is a discrete difference of lattice-point displacements u(r i ). It reduces to macroscopically varying engineering strains in the continuum limit of lattice constant a0 → 0, when the discrete differences become derivatives. Consider for simplicity, a cubic lattice with atoms at lattice points r i (scaled in a0 ) interacting through a two-body potential V that depends only on the magnitude of their separation r i,j = r i − r j . Then for a deformed lattice with points at Ri = r i + u(r i ), the energy change ∆V = V (|Ri,j |) − V (|r i,j |) is ∆V = V (|r i,j + u(r i ) − u(r j )|) − V (|r i,j |) .

(2.1)

For nearest-neighbour lattice points, r i = r j + µ ˆ (where µ ˆ=x ˆ, yˆ, zˆ) the discrete difference [46] of the displacement-vector components naturally appears, ˆ) − uν (r j ). More specifically, in changes of the squared ∆µ uν (r j ) ≡ uν (r j + µ separations (2.2) |Rj+ˆx,j |2 − |r j+ˆx,j |2 = 2Exx , the differences appear as 12 d(d + 1) components of a unit-cell, symmetric strain tensor that is defined by Eµν ≡

1 [∆µ uν (r i ) + ∆ν uµ (r i ) + Gµν ] . 2

(2.3)

8

S.R. Shenoy et al.

Here the last term Gµν ≡ ∆
µ u(r
i ) · ∆ν u(r i ) is the “geometric nonlinearity”. An expansion, ∆V = 12 µν i Kµν E 2 µν (r i )+· · · will include harmonic strain terms, with Kµν an appropriate spring-constant matrix. All this is simply to motivate the definition of the unit-cell strain tensor (2.3) of Level 1, that appears in the Ginzburg–Landau free energy expansion below. Although here only the static case is considered, the dynamic case has also been treated [37]. In a Lagrangian approach, expressing the displacement kinetic energy in terms of strain rates yields a strain-representation nonlinear dynamics for ferroelastic transitions; dynamic compatibility kernels can be of a resonant form to drive nonequilibrium interfaces at sound speeds [37]. The six physical strains that transform according to irreducible representations of the point group for the 3D cubic case, are one compressional, two deviatoric, and three shear strains, defined by linear combinations [38] 1 1 −1 e1 = √ (Exx +Eyy +Ezz ), e2 = √ (Exx −Eyy ), e3 = √ (Exx +Eyy −2Ezz ), 3 2 6 e4 = Eyz ,

e5 = Ezx ,

e6 = Exy .

(2.4)

The three physical strains in 2D for the square lattice are one compressional, one deviatoric and one shear strain 1 e1 = √ (Exx + Eyy ), 2

1 e2 = √ (Exx − Eyy ), 2

e3 = Exy ,

(2.5)

and we focus on this illustrative case, in Sect. 2.3.3. A simplifying approximation is useful. The displacement differences can be scaled in a typical strain value λ through ∆u → λ∆u or E → λE(λ), with λ determined in terms of material parameters, e.g. for FePd, λ ∼ 0.02 [37]. The last term in (2.3) is smaller by a relative factor of λ  1 and can be neglected as a first approximation, setting λGµν = 0. (The geometric nonlinearity can later be included perturbatively, or by using the second-order compatibility equations [45].) For now, we take strain as the defined linear symmetric tensor. A Ginzburg–Landau expansion in quadratic and higher invariants of λE(0) is then formally a power series in λ. 2.3.3 Strain Free Energy To describe structural transitions, with free energy nonlinearities to drive the finite-temperature transition, a partition function partial trace over Level 2 variables will yield a variational free energy Fstrain in terms of the Level 1 strains. The coefficient of the OP harmonic term, that decreases linearly on cooling is A(0) (T − Tc ), and the nonlinear terms generate a first-order phase transition, at some temperature T0 > Tc , that pre-empts the full softening on cooling to Tc . There are also non-OP harmonic terms. Scaling all strains in a typical value λ and all energy densities in a typical value D0 , and with scaled OP strains notationally distinguished through, e.g. ε ≡ e2 for a deviatoric OP,

2 Spin, Charge, and Lattice Coupling in Multiferroic Materials

9


the free energy Fstrain = D0 f . The dimensionless scaled free energy density f = f0 (ε) + fgrad (∆ε) + fnon (e1 , e3 ) can be expanded in symmetry-allowed invariants, that for the 2D square–rectangle case are 1 f = [(τ − 1)ε2 + ε2 (ε2 − 1)2 ] + ξ0 2 (∆ε)2 + (A1 e1 2 + A3 e3 2 ) . 2

(2.6)

Here the dimensionless ξ0 sets a domain-wall thickness. The scaling of the Landau free energy f0 (ε) to a factorized form as in the square brackets can be done for all ferroelastic transitions in 2D and 3D [37,47], generalizing a procedure of Barsch and Krumhansl [48]. The important dimensionless temperature variable τ = (T − Tc )/(T0 − Tc ) is unity at a temperature T0 , when all nonzero free energy minima in OP space fall on a “unit sphere” of radius |ε| = 1, and are degenerate with the ε = 0 high-temperature minimum. Dimensionless elastic parameters are elastic constant anisotropy ratios, related to the unscaled values by A1,3 = A(1,3) /[A(0) (T0 − Tc )], that can in principle be determined from experiment, molecular dynamics simulations, or electronic structure calculations. The scaled values can be large for anisotropic systems (e.g. for FePd parameters, A1 = A3 /2 = 150), that has consequences discussed later. It is tempting to say that since the non-OP free energy terms in (2.6) are harmonic, the optimization must set e1 = e3 = 0 at the outset, so that (for τ < 1) a uniform ε = ε¯(τ ) = 0 minimum determined by the Landau free energy alone, is picked out. However, this procedure incorrectly treats all strain components with wavevector k = 0 as independent, and misses nonuniform states that are also local minima. Simulations in a displacement representation of (2.6) minimize f in the d-dimensional variable u(r) at each lattice site. If instead, we want to work in terms of the physical unit-cell variables of compression, shear and deviatoric strain, then in fact, we have too many variables! The 12 d(d + 1) variables must be reduced in number by Nc = 12 d(d − 1) constraints, to yield the required number of 12 d(d + 1) − Nc = d degrees of freedom. Such constraints exist [34], and are implicit in the very definition (2.3) of the strain tensor. 2.3.4 St Venant Compatibility Constraints In a “dyadic” notation, the strain tensor in real and Fourier space (without geometric nonlinearity) can be written as  1 ∆u(r) + (∆u(r))T , 2 i E(k) = [Ku(k) + u(k)K] , 2 E=

(2.7)

where T is “transpose”, and in Fourier space, the difference operator is effectively ∆µ → i Kµ ≡ i 2 sin(kµ /2). (The differences are bond-midpoint variables [46], and the physical strains can be assigned to the unit-cell centres, i.e. the dual lattice.)

10

S.R. Shenoy et al.

Since Curl(Grad) acting on a single-valued function is zero, manifestly: Inc[E] ≡ ∆ × [∆ × E]T = 0 ; K × E × K = 0.

(2.8)

This compatibility constraint was obtained, in the continuum limit, by St Venant [49] in an 1864 commentary on a work of Navier, and its importance was emphasized more recently by Baus and Lovett [34]. It expresses the physical requirement that if a unit-cell is deformed, then the neighbouring and further-off cells must also deform in a decreasing way, so the cells all fit together in a smoothly “compatible” manner. Clearly, even if the cell that initiates these knock-on effects [1] is a pure deviatoric strain say, the compression/shear strains could still be needed, to maintain lattice integrity [37]. In (2.8), Inc[· · · ] defines an “incompatibility” operation on tensor fields, such as the Div[· · · ] or Curl[· · · ] operations on vector fields. In fact Inc[E] = 0 is a zero-incompatibility, no-defect condition analogous to the zero-divergence, no-monopole field equation Div[B] = 0 in electromagnetism. In terms of single-valued auxiliary variables, i.e. displacements or vector potentials u, A, both are trivial mathematical identities; but in terms of the physical variables, i.e. unit-cell strain or magnetic field E, B, both are physical field equations. In 1D there is only a single strain ε = ∆x ux (x) and there are no compatibility constraints: Nc = 12 d(d − 1) = 0. In 2D, the explicit difference equation with Nc = 1 is 1 1 √ ∆2 e1 − 2∆x ∆y e3 − √ (∆2x − ∆2y )e2 = 0 . (2.9) 2 2 In 3D with Nc = 3 we have [38, 45] 2∆y ∆z Eyz = ∆2z Eyy + ∆2y Ezz , 2∆z ∆x Ezx = ∆2x Ezz + ∆2z Exx , 2∆x ∆y Exy = ∆2y Exx + ∆2x Eyy .

(2.10)

There are also three other equations that emerge, ∆y ∆z Exx + ∆2x Eyz = ∆z ∆x Exy + ∆x ∆y Ezx , ∆z ∆x Eyy + ∆2y Ezx = ∆y ∆z Exy + ∆x ∆y Eyz ,

(2.11)

∆x ∆y Ezz + ∆2z Exy = ∆y ∆z Ezx + ∆z ∆x Eyz , but solving for shears in Fourier space, or applying appropriate difference operators, one gets identities. 2.3.5 Compatibility Potentials The derivations of the St Venant compatibility potentials for a given transition are straightforward, as the interested reader can easily confirm by working through the square–rectangle case below. (The uninterested one can skip directly to (2.14), and conceptual comments.)

2 Spin, Charge, and Lattice Coupling in Multiferroic Materials

11

There are 12 d(d + 1) physical strains, of which NOP are order parameters, and the other n = 12 d(d + 1) − NOP components are non-OP strains. One can use Nc Lagrange multipliers for the constraints [35–37]. More simply [42]: 1. Solve the Nc compatibility equations for n non-OP strains in Fourier space and substitute into the harmonic non-OP free energy, thus guaranteeing compatibility. The free energy now has d = (d/2)(d + 1) − Nc strains. 2. Then further minimize (freely) with respect to the remaining n − Nc = d − NOP
0 non-OP strains, that are harmonic or bilinear in the free energy. This procedure converts the harmonic, local, non-OP free energy fnon to a harmonic, non-local OP free energy fcompat . For the square–rectangle transition, NOP = 1, n = 2 and from (2.9), the Nc = 1 compatibility equation in fourier space is e1 (k) = −[O2 (k) ε(k) + O3 (k) e3 (k)]/O1 (k) ,

(2.12)

where O1 (k) = − √12 |K|2 , O2 (k) = √12 (Kx2 − Ky2 ), O3 (k) = 2Kx Ky . By substituting for e1 and minimizing in e3 , the final result yields the compressional and shear strains as proportional to the OP deviatoric strain, e1,3 (k) = B1,3 (k) ε(k). Here B1 ≡ −O1 O2 /D, B3 ≡ −(A1 /A3 ) O3 O2 /D, whereas D ≡ [O12 + (A1 /A3 ) O32 ], and it is seen that the compatibility constraint (2.12) is satisfied as an identity. Note that for an equilibrium ε(k) texture peaking at diagonal orientations kˆx = kˆy , the non-OP strains e1,3 vanish, as the coefficients B1,3 ∼ O2 = 0. The non-OP harmonic free energy density becomes [35–37, 42] fnon (e1 , e3 ) = fcompat (ε) =

A1 A3 A1 |e1 |2 + |e3 |2 = U (k) |ε(k)|2 , 2 2 2

(2.13)

where U (k) = |B1 |2 + (A3 /A1 )|B3 |2 . The free energy Fstrain (ε) is obtained in terms of the OP alone, with the non-OP strains as derived quantities  A1  [(τ − 1)ε2 + ε2 (ε2 − 1)2 ] + fgrad (∆ε) + ε(r)ε(r
) U (r − r
) . Fstrain = 2
r r,r

(2.14) This is a central conceptual result. In the usual displacement representation, every term in the free energy (2.6) involves anisotropic, nearest-neighbour differences of a d-dimensional variable u defined at the lattice points, e.g. 6 e21 = (1/2)(∆x ux + ∆y uy )2 , while 6 = (1/8)(∆x ux − ∆y uy ) . On the other hand, in the OP strain representation, the free energy of (2.14), in the spirit of Landau, depends only on the NOP ≤ d order parameters (on the dual lattice), and there is now an explicit, far-neighbour, potential that carries all the anisotropies [36, 37]. The Landau part in square brackets is diagonal in coordinate space, while the compatibility potential is diagonal in Fourier space. The St Venant kernel U (k) = O22 /2D ∼ (Kx2 − Ky2 )2 /K 2 , and the lattice Green’s function

12

S.R. Shenoy et al.

G(k) = [Kx2 + Ky2 ]−1 , both reflect the discreteness of the lattice in their ˆ anisotropy at short wavelengths. However, at long wavelengths U (k) → U (k) is still anisotropic and moreover, independent of the wavevector magnitude, in contrast to the isotropic G(k) → 1/k 2 that depends (only) on |k|. This generic feature of U arises from solving the compatibility equation in terms of factors Oi /O1 (i = 2, 3), that depend on ratios Kµ2 /K 2 → kµ2 /k 2 = kˆµ2 . As an important consequence, U (R) is asymptotically anisotropic, with a numerator that reflects the symmetry of the unit cell multiplying a power-law decay whose exponent is the dimensionality. By contrast, G(R) ∼ 1/Rd−2 is asymptotically isotropic. Explicitly, the d = 2 square-lattice kernel in coordinate space, at distances beyond a lattice scale is [35–37]  ˆ ∼ cos(4[θ − θ
])/Rd , (2.15) U (R) → dd k eik·R U (k) where rˆ · rˆ
= cos(θ−θ
) and the fourfold symmetry is manifest. Clearly, the asymptotic power-law decay has nothing to do with proximity to a second-order critical point: it arises from a change of variables (k → k/R) and wavevector phase space, with no contribution from the |k|-independent, long wavelength kernel. The strain at a site can receive sign-varying, conflicting instructions from the other strains around it, in elastic frustration. While the spatial integral of the truly long-range G(R) diverges as a power of system size, and that of an isotropic power-law 1/Rd is logarithmically divergent, the integral is zero over the sign-varying U (R) power-law potential. The compatibility kernels are different for different ferroelastic transitions, as they encode the transition symmetries. They have been evaluated for the case of 2D transitions, including square–rectangle (NOP = 1, NV = 2), triangle to centered rectangle (NOP = 2, NV = 3), triangle-oblique (NOP = 2, NV = 6) [37]. For 3D transitions, evaluated kernels include cubic–tetragonal (NOP = 2, NV = 3), cubic–trigonal (NOP = 3, NV = 4) and so on [40, 47]. 2.3.6 Other Approaches There are many other complementary approaches to elastic textures, and we can merely list some of them. 1. Simulations in the displacement representation: This is widely used [35, 50]. Strains are derived from the equilibrium displacement-vector texture. 2. Sharp-interface minimizers: In the limit ξ0 → 0 the gradient cost tends to zero, yielding sharp-step domain walls. The applied mathematics problem is then to find distributions of domain walls with positive/negative constant-strain regions in between them, such that the nonlinear free energy is a local minimum [51, 52]. 3. Simulations in morphological variables: This “phase-field” method focuses on atomic-basis morphological variables coupled to the strain, to describe the variants, with the inhomogeneous strain treated as a local-inclusion.

2 Spin, Charge, and Lattice Coupling in Multiferroic Materials

13

The effective free energy is nonlinear in the morphological variables, and has an anisotropic power-law potential between their squares. Strains are derived quantities [53]. Nonperiodic or disordered elastic systems, such as network glasses, can be studied through a discrete approach to their rigidity, that finds floppy, intermediate and rigid phases, as the average number of local bonding constraints varies [54]. In contrast to the above approaches, we describe here a representation that directly works with the strain, using derived compatibility kernels specific to each transition and an anisotropic power-law between OP strains (not their squares). The displacement vector can then be obtained from the strain field [34]. This approach is easily extended to include strain coupling to other variables, as discussed in the following sections. Complementary models for charge, spin or lattice texturings include those that invoke Coulombic forces and/or polaronic lattice couplings; charge states as Eshelby inclusions; quenched-disorder Griffiths models with locally suppressed transition temperatures; Ising-spin toy models; or inhomogeneities linked to trapped stresses [6, 55, 56].

2.4 Inhomogeneities in Ferroelastics The local minima of the free energy Fstrain of (2.14) are found by longtime ε(r, t) states emerging from a relaxational dynamics (that can be obtained in some limit from an underdamped dynamics), ε˙ = −∂Fstrain (ε)/∂ε

(2.16)

with random ε(r, t = 0) initial conditions. The non-OP strains e1 , e3 are obtained as derived quantities. With additional free energy terms, as discussed below, the asymptotic states can show equal-width twins, and tweed. Twins Figure 2.2a shows, for large anisotropies A1 = 100, the final state of oriented domain walls. The preferred orientation follows from the compatibility kernel U ∼ (kˆx2 − kˆy2 )2 that favours diagonal domain walls, but does not set a mesoscale domain wall separation length. The costly non-OP strains e1,3 (k) vanish at diagonal orientations, as mentioned, in a kind of “elastic Meissner effect” [36]. (This is not true for all textures – e.g. the triangle to centred rectangle transition has states with three degenerate structural domains meeting at a point, where compatibility forces the e1 (r) non-OP strain to be nonzero [37].) Figure 2.2b shows that for smaller A1 1 and other parameters the same, we get multiscale domains. These are qualitatively similar to martensites, that Krumhansl has termed a paradigm for the multiscale concept [10].

14

S.R. Shenoy et al.

Fig. 2.2. Ferroelastic domain walls, below the square-to-rectangle transition temperature. (a) Large coefficient of the (bulk) compatibility potentials A1 = 100
1. (b) Smaller coefficient of the (bulk) compatibility potentials, A1 ∼ 1. (c) Same as (a), but with different initial random seed, showing different quasidegenerate final state. (d) Same as (a), but with an added surface compatibility potential setting a length scale for equal-width twins

Why do such domain-wall states occur? After all, since the coefficient ξ02 of the squared-gradient term in the free energy is positive, domain walls always cost energy. Figure 2.2c shows that with the same parameters as Fig. 2.2a, but a different initial-state random seed, the number Ndom of domain walls changes; although with the free energy density f = F/L20 almost unchanged. This is easily understood: if the uniform case ε = ε¯(τ ) has a free energy density f = −|f0 (¯  of width ξ0 will have a nearly ε)|, then a state with walls ε)| − (ξ0 /L0 )Ndom ε¯2 . And once an elastic soliton (or degenerate f = − |f0 (¯ domain wall) is oriented by the compatibility potential A1 U (that vanishes at a preferred angle), then the free energy barrier to a deviating segment of finite-fraction L0 length will be Fbarrier ∼ A1 (ξ0 L0 ) ε¯2 . Thus one has the possibility of multiple wells separated by macroscopic barriers, in a rugged “free energy landscape” sculpted by the St Venant lock-in of multiple domainwall segments, with all preferred orientations present. It would be interesting to make this “configurational strain-glass” picture more quantitative [57]. The Ndom = 0 uniform state may be the lowest in energy, but only marginally so; moreover it has many competitors, and so is hard to find! Of course, all this is with periodic boundary conditions. However, finite systems or nonperiodic boundary conditions can also be handled in principle. Consider an L0 × L0 square martensite embryo within a sea of austenite.

2 Spin, Charge, and Lattice Coupling in Multiferroic Materials

15

The austenite/martensite interface (habit plane) at x = 0 say, can be treated by including decaying surface strains ε˜(r), in the austenite; and also adding on to a “bulk” strain ε¯(r) deep inside the martensite. We demand continuity of the total strain on either side, plus strain–gradient matching as dictated by integrating the St Venant equation across the boundary. Then the surface strain amplitudes [in a Laplace–Fourier expansion for ε˜(r)] can be written in terms of the bulk amplitudes [in a Fourier expansion for ε¯(r)]. The simulations can be performed in terms of the periodic boundary condition bulk strains in an L0 × L0 surrogate system, with the surface strain contributions integrated out to yield a surface compatibility potential that adds on to the bulk compatibility potential. The necessarily nonzero decay constant in the positive x direction in the austenite side, then forces a modulation wavevector or martensitic twinning in the y direction inside the
embryo. Explicitly, ε(kx , ky )|2 /|ky |, in this elastic fringing-field energy is Fsurf ∼ (1/L0 )A3 k |¯ agreement with a displacement representation calculation. Figure 2.2d shows that Fsurf clearly does set a mesoscopic length scale (has long wavelength |k| dependence), yielding equal-width “true” twins, and inducing a width–length scaling W ∼ L0 1/2 , as in early experiments [27, 36]. Tweed In (some) martensites above the structural phase transition, nanoscale crisscross domains are seen, that have been called “tweed” textures by Tanner [28]. Tweed (and twins) has been seen in cuprates, giant magnetoresistance (GMR) manganites, and relaxor ferroelectrics [7, 31, 32]; and magnetic tweed can also be imaged and modelled [58]. Its ubiquity suggests an explanation in terms of a common variable, e.g. strain. For ferroelastics, tweed has been linked [28] to an incipient softening on cooling of the elastic constant coefficient C
= C11 − C12 of the harmonic deviatoric strain ε2 , that is τ in scaled form. Since compositional disorder is almost always present in alloys (and the transition temperature varies sharply with it), models that couple strains to the local composition η(r) have been considered. These ftweed terms include: (i) η(r)ε2 (r) or spatially local structural transition temperature [35]; (ii) η(r)e1 (r) or local compressional stress [59]; (iii) [(∆2x − ∆2y )η(r)]ε(r) or local deviatoric stress [36]. Simulations with a fixed disorder realization show diagonally oriented tweed-like domains. For model (ii), an infinite anisotropy (A1 → ∞) yields a spin-glass picture of tweed [7, 35]. If η(r) is treated like an annealed disorder random variable with distrib2 2 ution P = e−η /2σ in the partition function, then the average of the Boltzmann factor e−βFtweed for models (i) and (ii) yields just renormalization of the ε4 or e1 2 terms, respectively. However, model (iii) yields the nontrivial term fcross ∼ −βσ 2 [(∆2x − ∆2y )ε]2 , that favours crossings/steps in the domain-walls (being nonzero at intersections/corners). Since η(r) is actually a quencheddisorder random variable, averages must be performed through replicas and the n → 0 trick. Then including only Gaussian replica correlations, we obtain

16

S.R. Shenoy et al.

the effective tweed term ftweed = [1 − (Ttweed /T )2 ]fcross with an extra temperature factor, that favours tweed only at temperatures T > Ttweed ∼ σ.

2.5 Inhomogeneities in Multiferroic Oxides High-resolution microscopies on complex oxides have revealed multiscale heterogeneities on scales from hundreds to a few lattice spacings. These include stripes, conducting filaments, superconducting islands and so on [4–6]. Cuprates like LaSrCuO undergo tetragonal to orthorhombic ferroelastic transitions with strain as the primary order parameter, and a N´eel transition of the antiferromagnetic (AF) state at lower temperatures. Both structural and AF transition temperatures vary sharply with doping, e.g. substitution of one rare earth element by another of a different valency. The superconducting and pseudogap temperatures show large isotope effects over a range of dopings [60]; and there can be a downward cusp close to 1/8 doping. Angleresolved photoemission shows lattice signatures [61]. Shape memory effects have been reported [17]. Diffuse X-ray, and neutron scattering show stripes of competing unit-cell structures at nanoscales, and staggered-magnetization domain walls [21, 22]. Scanning tunneling microscopy (STM) in the superconducting state shows a gap magnitude that varies over regions from 1 nm upwards [20]. Remarkably, a uniform magnetic field can couple to mesoscale elastic twin boundaries, and rotate their orientation [18]. Hence it is clear that cuprates can show both cross-variable and cross-scale response, and are complex multiferroics with strong charge/spin/strain coupling. Colossal magnetoresistance (CMR) manganites like LaCaMnO show a complex phase diagram as a function of doping, with the transition temperatures between phases showing cusps close to multiples of 1/8 [3, 6]. At high and low dopings, there are AF insulator, and charge/orbital ordered regions. At intermediate dopings there is a ferromagnetic metal region that goes over into a paramagnetic insulator as the temperature is raised, with the Curie and metal–insulator temperatures quite close, and a resistivity peak around these temperatures. The Curie temperature has a strong isotope effect [60]. Theoretical double exchange models link the electron mobilities to orderings of the core level ferromagnetic spins [3]. With resistance changing by orders of magnitude, the “colossal” cross-variable responses include magneto-resistance in a magnetic field ∼ T [3]; stresso-resistance under external pressure ∼ GPa [13]; nonlinear resistance with switching/memory, under a voltage ∼ kV [14]; and photoconductance under laser power ∼ mW [62]. In short, here too, there is strong charge/spin/strain coupling and complex multiferroic behaviour. STM probe studies by Faeth et al. [19] on CMR manganites reveal that the above macroscopic functionalities conceal a hidden nanoworld of spatial variations. Although STM is a surface probe, they found that the integrated STM resistivity versus temperature closely tracks the measured bulk resistivity peak, implying the intrinsic inhomogeneity is not just superficial.

2 Spin, Charge, and Lattice Coupling in Multiferroic Materials

17

Fig. 2.3. STM picture in coordinate space of manganite local conductance near the ferromagnetic transition [19]. Insulator/metal denoted as light/dark. Panels, reading left to right from the top, show cross-scale responses in meso- and nanoscale texturing, with increasing uniform magnetic field in Tesla: 0, 0.3; 1.0, 3.0; 5.0, 9.0. Scale bar in the lower-most right panel (9 T) is 100 nm

Figure 2.3 shows the local conductance at a few volts. The dark (light) areas are metallic (insulating). The upper left panel shows that the zero field resistivity is already intrinsically inhomogeneous, from nano- to mesoscales. With increasing strength of magnetic field, the meandering high-conductance channels expand, and the resistivity drops: the global CMR is associated with the magnetic field or temperature controlling a percolative metallicity, that adaptively coexists with insulating regions [26]. This is a remarkable cross-scale response, as it is uniform fields that produce these nanoscale, “cloud-like” conductance variations [19]. It would be interesting to see how STM images vary with applied uniform voltage, stress, or illumination. Analogies between manganites and martensites [10] have been noted in lenticular domain-wall texturings; or in their dynamics, with long incubation times, or avalanche-type acoustic emissions; and noise spectroscopy can be a state diagnostic [25, 29, 56, 63]. Multiferroics with a ferroelectric-polarization order parameter have not been emphasized here, but they can also show tweed and twins [32]. Piezoresponse force microscopies have revealed meandering mesoscale channels [64] reminiscent of manganites [19]. Magnetic fields can switch electric polarization in TbMnO ferroelectric manganites [15]. Early work has indicated ferroelectric signatures in cuprates [65] (and there are Level 2 models for pairing based on paraelectric fluctuations [66]).

18

S.R. Shenoy et al.

With the nonlocality already inherent in the anisotropic and unruptured lattice, the viewpoint is that these complex textures are the adaptive responses of nonlinear media to simple local couplings of strain to other variables.

2.6 Charge and Spin as Local Stresses and Transition Temperatures 2.6.1 Coupling of Variables The charges and spins couple in symmetry-allowed ways to strain tensor components. When coupled to linear (or quadratic) powers of strains, they act as local stresses (or local elastic constants). Conversely, when strains are coupled to quadratic powers of the magnetization (whether staggered or direct), then they act as local magnetic transition temperatures (whether Curie or N´eel). Compatibility-induced power-law correlations then spread the influences of charge and spin states over large distances, and nonlinearities can mix scales. The magnetization contribution is (with ξm being the magnetic domain wall width)    1 2 ξm (∆m)2 , (2.17) (T − Tcm )m2 + m4 + Fmag = g0 2 r r where g0 < 1 is a ratio of magnetic to elastic energy densities. The charge carrier–lattice interaction is in general, a symmetry-allowed coupling to a specific physical strain. For harmonic, isotropic solids, it is often sufficient to write displacement as u(q) = u ˆ u(q) where u ˆ is a polarization direction, and then angular average the coupling to α(q) ≡ α to get an “electron–phonon” inter
action, −α q n(q) u(q). However for anisotropic multiferroics, nonlinearities single out the OP strains; and spins/charges have different symmetry-dictated couplings to the different physical strains, that have unequal elastic constants. It is preferable to remain in the strain representation, and avoid truncations. A charge state in a directional d-wave orbital can act as a deviatoric stress, while an isotropic “mobile” [67] charge state can act as a compressional stress. To focus on the anisotropy from the compatibility potentials, we consider an s-wave number density n(|r|) > 0 of the charge. Ignoring higher order magnetoelastic terms such as Amε m2 ε2 , the charge/spin/strain couplings [39]  An1 ne1 + Am1 m2 e1 + Anε nε2 + Anm nm2 (2.18) Fcoupling = r

act on strains as local stress “p1 e1 ” terms, or local shifts in the structural transition temperature “τ ε2 ”. Note that the m2 terms can be combined with the (T − Tcm )m2 term of (2.17) to define a Curie or N´eel local transition temperature, enhanced/suppressed essentially by the annealed local strain
Tcm (r) ≡ Tcm − [Am1 e1 (r) + Anm n(r)]/g0 .

(2.19)

2 Spin, Charge, and Lattice Coupling in Multiferroic Materials

19

The signs of the coupling constants are chosen to mimic some aspects of cuprates and manganites [39]. Thus Am1 > 0 implies that compression e1 < 0, that brings spins closer, favours magnetization, where m is the direct (staggered) magnetization for manganites (cuprates). Since doping reduces the structural transition temperature, Anε > 0, and for cuprates Anε  1. For manganites we take Anm < 0 to mimic double exchange, where the presence of a mobile hole favours local ferromagnetism m(r) of the core spins. For cuprates, Anm > 0, as doping dilutes spins, and in fact strongly suppresses the N´eel temperature, Anm  1. 2.6.2 Simulations We use a coupled relaxational dynamics with random initial conditions in strain and magnetization, to find the free energy local minima through m ˙ = −∂F/∂m;

ε˙ = −∂F/∂ε ,

(2.20)

where we drop Ftweed and Fsurf of Sect. 2.4 for simplicity: F = F (ε, n, m) = Fstrain + Fmag + Fcoupling .

(2.21)

The number-densities of the charges are normalized exponentials ∼e−2r , placed randomly, with fixed positions and profiles. It would be interesting to introduce a charge-density (ordering) dynamics, as considered elsewhere [68]. The bare Curie (N´eel) transition temperatures are taken to be Tcm = 0 (Tcm = 0) for manganite (cuprate) parent compounds treated as paramagnets 2 = 12 , T0 = (antiferromagnets). We take [39] A1 = 50 = A3 /2, ξ02 = 18 , ξm 1, Tc = 0.8, g0 = 0.3, Am1 = +5. The other, “manganite” parameters are Tcm = 0, Anm = −1, An1 = +5, Anε = +2. The “cuprate” parameters are Tcm = 0.6, Anm = +9, An1 = −5, Anε = 20. This is a regime of globally weak magnetism, relatively strong lattice couplings, and dominant compatibility forces. Figure 2.4a shows the e1 (r) strain of two textured manganite polarons, that are large and floppy, and deform each other [39]. These are multivariable polaronic elasto-magnetic textures or pemtons, and finite-doping textures are multipemton states. Note that although the local stress An1 n has a given sign, compatibility induces quadrupoles, with both signs of strain (compression and expansion). Figure 2.4b shows that x = 0.15 produces large m2 = 0 magnetization variations at multiple scales, but only in those regions where the temperature is below the strain-induced local transition tempera

(r); whereas m2 = 0 in T > Tcm (r) regions. The ferromagnetic ture, T < Tcm formation of a symmetry-breaking average m = 0 on this (conducting) backbone will also have a percolative character. For cuprates Fig. 2.4c shows a more compact, smaller pemton, while Fig. 2.4d shows stripe-like structural variants for x = 0.1. The multipemton strain structure factors |ε(k)|2 , |e1 (k)|2 reveal the fourfold symmetry hidden in the apparently disordered coordinate-space

20

S.R. Shenoy et al.

Fig. 2.4. “Manganite” parameters: see text. (a) Two charges showing compressional strain of two textured polarons. (b) Multipolaron magnetization for random-position doping fraction x = 0.15. “Cuprate” parameters: see text. (c) Single textured polaron. (d) Multipolaron deviatoric strain for doping fraction x = 0.1 showing stripelike texturing

textures [39,42], showing butterfly-shape contours reminiscent of experimental diffuse scattering plots [69]. The application of long-wavelength external magnetic fields and stresses produces delicate short-wavelength “cloud-like” variations [39]. Note that in linear response, e.g. m(q, ω) = χ(q, ω)h(q, ω), the response wavevector/ frequency is the same as that of the applied field, and increasing the field strength merely magnifies the response. However, very different modes can 6 be mixed by the nonlinear free energy strain ε6 (r, t) → i=1 ε(qi , ωi ) and 4 magnetization m4 (r, t) → i=1 m(qi , ωi ) terms. Hence cross-scale responses in length and time are another signature of nonlinearity [19, 44]. Thus simulations show multivariable, multiscale texturing, with textured single polarons, and stripe or channel-like multipolaronic states. There is one conceptual problem, however. The textures are supposedly from the generic feature that doping does not produce dislocations. But then complex textures

2 Spin, Charge, and Lattice Coupling in Multiferroic Materials

21

should always be seen – the picture is “too generic”! A further analysis [70] shows that the key response function is the compliance describing the strain from local internal stresses, e1 (k) = χ11 (k) p1 (k). It contains the compatibility kernel in the denominator, and has a power-law √ tail in coordinate space, χ11 (R) ∼ 1/Rd , beyond an onset length ξ = ξ0 / A1 . Hence if ξ  1, the compliance is already negligible before the power-law sets in; while for materials with ξ ∼ 1 there is power-law behaviour over most of the range. For FePd, we have ξ = 1.2. Although we have emphasized annealed-variable adaptive texturing, it is clear that quenched variables play an important and complementary role. Quenched compositional disorder acting as local stresses at random sites will have nonlocal compliance effects on the strain even at far-off sites, with total random-stress correlations showing APL decays. This could additionally motivate effective random-field, correlated disorder models as studied by Dagotto and coworkers [6]; or nonequilibrium stress-trapping scenarios of inhomogeneities [56]. It would be interesting to extend various Level 2, models for CMR behaviour or cuprate pairing [66, 71] to include Level 1 strain compatibility potentials and annealed-variable/quenched-disorder internal stresses, that induce multivariable heterogeneity. Parameter regimes where electronic states are sensitive to local quenched disorder [72], could be relevant for such straininduced wavefunction texturing. The inhomogeneous states will have their own electronic and lattice excitations, e.g. domain-wall ripplings (“dyadons”) [27].

2.7 Summary and Further Work Intrinsic heterogeneities in multiferroics are both multivariable and multiscale, and can have cross-variable and cross-scale responses. As noted by Krumhansl, the simultaneous presence of multiple, distinct scales could require novel approaches to materials science that treat Ginzburg–Landau functionals as effective field theories [10]. Although there are different levels of description as in Fig. 2.1, the viewpoint outlined here regards the unit-cell strain tensor as a central factor in multiferroic texturings, that involve three elements: symmetry, compatibility, and nonlinearity. The symmetry of the discrete point group of the unit-cell dictates the series expansion of the strain free energy in allowed invariant combinations of physical strain-tensor components. It also determines allowed local couplings of strain components to other variables like charge density, magnetization, and electric polarization, that can induce strain-driven local variations of (other) multiferroic transition temperatures. Compatibility constraints on strain components become effective interaction potentials with (symmetryspecific) sign-varying anisotropies and (dimensional) power-law decays, when some strains are eliminated. And the coupled order-parameter nonlinearity that is present over wide temperature ranges supports solitonic domain walls

22

S.R. Shenoy et al.

(elastic or otherwise), that under orienting compatibility potentials, can lock in to complex patterns, as local minima of a multivalley free energy landscape: a multiferroic configurational glass. Further texturing, or a stabilization of inhomogeneous states relative to uniform states, could occur from compositional quenched disorder, that is nonlocally magnified by compatibility effects. A microscopic treatment of the ideas would clearly be useful. If we integrate out the strains in Gaussian approximation, the
effective free energy for the remaining variables in Fourier space Feff = D0 k feff is

1 feff = ACoul |n|2 /k 2 + χ11 |An1 n + Am1 m2 |2 + · · · , 2

(2.22)

where we have added a charge–charge Coulomb interaction, and have not shown the implicit higher-order in n–n, n–m2 , and m2 –m2 interactions, from strain nonlinearities. Even for large isotropic Coulomb strength ACoul  (An1 , Am1 ), the treatment of correlated electrons [73] could more generally include these qualitatively distinct anisotropic interactions between charges and spins; and the consequent intrinsic inhomogeneities. One might conjecture that whereas Mott and Anderson localizations involve metal-insulator transitions across the system, there is a possibility here of “St Venant localization”, i.e. inhomogeneous multipolaronic trapping, that coexists with the metallicity of electrons self-focussed into channels, as they propagate through a nonlinear, anisotropic, elastic medium, with mobility edges in coordinate space. The emerging viewpoint is speculative and undoubtedly needs modification in detail, but seems worth exploring further. The unifying concept is of annealed order parameters that are adaptively complex in the sense of Krumhansl, whose insights have helped bring conceptual order to so many issues in condensed matter physics and materials science. Acknowledgements It is a pleasure to thank A.R. Bishop for fruitful collaborations. Partial support of this work is gratefully acknowledged, by SRS from ICTP, Trieste; and TL, AS from the US Department of Energy.

References 1. E.K.H. Salje: Phase Transformations in Ferroelastic and Coelastic Solids (Cambridge University Press, Cambridge, UK, 1990) 2. Z. Nishiyama: Martensitic Transformations (Academic, New York, 1978) 3. B. Raveau, C.N.R. Rao (Eds.): Colossal Magnetoresistance and Related Properties (World Scientific, Singapore, 1998); A.J. Millis: Nature 392, 147 (1998) 4. A.R. Bishop, S.R. Shenoy, S. Sridhar (Eds.): Intrinsic Multiscale Structure and Dynamics in Complex Electronic Oxides (World Scientific, Singapore, 2003)

2 Spin, Charge, and Lattice Coupling in Multiferroic Materials

23

5. A. Bianconi (Ed.): Symmetry and Heterogeneity in High Temperature Superconductors (Plenum, New York, 2004) 6. E. Dagotto (Ed.): Nanoscale Phase Separation and Colossal Magnetoresistance (Springer, Berlin Heidelberg New York, 2003); J. Burgy, A. Moreo, E. Dagotto: Phys. Rev. Lett. 92, 097202 (2004) 7. Y. Bar-Yam, T. Egami, J. Mustre-de Leon, A.R. Bishop (Eds.): Lattice Effects in High-Tc Superconductors (World Scientific, Singapore, 1992) 8. E.K.H. Salje, A.S. Alexandrov, W.Y. Liang: Polarons and Bipolarons in High Tc Superconductors (Cambridge University Press, Cambridge, UK, 1995) 9. V.K. Wadhawan: Introduction to Ferroic Materials (Gordon and Breach, Amsterdam, 2000) 10. J.A. Krumhansl: Mater. Sci. Forum 327–328, 1 (2000) 11. E.F. Wassermann, M. Acet: this volume 12. Y. Murakami, D. Shindo, K. Oikawa, R. Kainuma, K. Ishida: Acta Mater. 50, 2173 (2002); A.N. Bogdanov, A. deSimone, S. Muller, U. Rossler: J. Mag. Magn. Mater. 265, 204 (2003) 13. H. Huang, T. Palstra, S.W. Cheong, B. Batlogg: Phys. Rev. B 52, 15046 (1995) 14. S.B. Ogale, V. Talyansky, C.H. Chen, R. Ramesh, R.L. Greene, T. Venkatesan: Phys. Rev. Lett. 77, 1159 (1996) 15. T. Kimura, T. Goto, N. Shintani, K. Ishizaka, T. Arima, Y. Tokura: Nature 426, 55 (2003); H. Hur, S. Park, P.A. Sharma, J.S. Ahn, S. Guha, S.-W. Cheong: Nature 429, 392 (2004) 16. P.H. Hor, L. Gao, R.L. Meng, Z.J. Huang, Y.Q. Wang, K. Forster, J. Vassilious, C.W. Chu, M.K. Wu, J.R. Ashburn, C.J. Torng: Phys. Rev. Lett. 58, 911 (1987) 17. R. Tiwari, V.K. Wadhawan: Phase Transitions 35, 47 (1991) 18. A.N. Lavrov, S. Komiya, Y. Ando: Nature 418, 385 (2002) 19. M. Faeth, S. Friesen, A.A. Menovsky, Y. Tomioka, J. Aarts, J.A. Mydosh: Science 285, 1540 (1999) 20. K.M. Lang, V. Madhavan, J.E. Hoffman, E.W. Hudson, H. Eisaki, S. Uchida, J.C. Davis: Nature 415, 412 (2002) 21. A. Bianconi, N.L. Saini, A. Lanzara, M. Missori, T. Rosetti, H. Oyanagi, H. Yamaguchi, K. Oka, T. Ito: Phys. Rev. Lett. 76, 3412 (1996) 22. J.M. Tranquada, B.J. Sternleich, J.D. Axe, Y. Nakamura, S. Uchida: Nature 375, 561 (1995) 23. T. Egami: this volume 24. S.J.L. Billinge in [4]; E.S. Bozin, G.H. Kwei, H. Takagi, S.J.L. Billinge: Phys. Rev. Lett. 84, 5856 (2000) 25. V. Podzorov, B.G. Kim, V. Kiryukhin, M.E. Gershenon, S.-W. Cheong: Phys. Rev. B (Rapid Commun.) 64, 140406 (2001); V. Hardy, S. Majumdar, S.J. Crowe, M.R. Lees, D.McK. Paul, L. Herv´e, A. Maignan, S. H´ebert, C. Martin, C. Yaicle, M. Hervieu, B. Raveau: Phys. Rev. B 69, 020407 (2004) 26. C. Renner, G. Aeppli, B.G. Kim, Y.A. Soh, S.W. Cheong: Nature 416, 518 (2002) 27. B. Horowitz, G.R. Barsch, J.A. Krumhansl: Phys. Rev. B 43, 1021 (1991); Phys. Rev. B 36, 8895 (1987) 28. L.E. Tanner, A.R. Pelton, R. Gronsky: J. de Physique (Paris) Colloque C4, 169 (1982); T. Cast´ an, A. Planes, A. Saxena: Phys. Rev. B 67, 134113 (2003) 29. E. Vives, J. Ort´ın, L. Ma˜ nosa, I. R` afols, R. P´erez-Magran´e, A. Planes: Phys. Rev. Lett. 72, 1694 (1994); T. C´ astan, E. Vives, L. Ma˜ nosa, A. Planes, A. Saxena: this volume

24

S.R. Shenoy et al.

30. V.K. Wadhawan: Phys. Rev. B 38, 8936 (1988) 31. L. Yiping, A. Murthy, G.C. Hadjipayanis, H. Wan: Phys. Rev. B 54, 3033 (1996) 32. O. Tikhomirov, H. Jiang, J. Levy: Phys. Rev. Lett. 89, 147601 (2002); S.P. Venkateswaran, M. De Graef: this volume 33. M. Seul, D. Andelman: Science 267, 476 (1995) 34. M. Baus, R. Lovett: Phys. Rev. Lett. 65, 1781 (1990); Phys. Rev. A 44, 1211 (1991) 35. S. Kartha, J.A. Krumhansl, J.P. Sethna, L. Wickham: Phys. Rev. B 52, 803 (1995) 36. S.R. Shenoy, T. Lookman, A. Saxena, A.R. Bishop: Phys. Rev. B 60, 12537 (1999) 37. T. Lookman, S.R. Shenoy, K.Ø. Rasmussen, A. Saxena, A.R. Bishop: Phys. Rev. B 67, 024114 (2003); D.M. Hatch, T. Lookman, A. Saxena, S.R. Shenoy: Phys. Rev. B 68, 104105 (2003) 38. A. Saxena, T. Lookman, A.R. Bishop and S.R. Shenoy in [4] 39. A.R. Bishop, T. Lookman, A. Saxena, S.R. Shenoy: Europhys. Lett. 63, 289 (2003); cond-mat/0304198 40. K.Ø. Rasmussen, T. Lookman, A. Saxena, A.R. Bishop, R.C. Albers, S.R. Shenoy: Phys. Rev. Lett. 87, 055704 (2001) 41. T. Lookman, A. Saxena, A.R. Bishop, S.R. Shenoy: J. Phys. (Paris) Coll. IV 112, 195 (2003) 42. S.R. Shenoy, T. Lookman, A. Saxena and A.R. Bishop, in [4, 5] 43. P. Chaikin, T. Lubensky: Principles of Condensed Matter Physics. 2nd edn. (Cambridge University Press, Cambridge, UK, 2000) 44. S.N. Behera, A. Khare: Pramana (Bangalore) 15, 245 (1980); F. Falk: Z. Phys. B 51, 177 (1983); A. Bussmann-Holder, A.R. Bishop: J. Phys.: Condens. Matter 16, L313 (2004) 45. S.F. Borg: Fundamentals of Engineering Elasticity, 2nd edn. (World Scientific, Singapore, 1990) 46. B.H. Chirgwin, C. Plumpton: A Course of Mathematics for Scientists and Engineers. Vol. 1, 2nd edn. (Pergamon, London, 1970), p. 321 47. S.R. Shenoy, T. Lookman, A. Saxena: unpublished 48. G.R. Barsch, J.A. Krumhansl: Metallurg. Trans. A 18, 761 (1988); Proc. ICOMAT-92 (1992) (Monterey Inst. Adv. Stud., Monterey, CA, 1993), p. 53 49. http://www-gap.dcs.st-and.ac.uk/∼ history/Mathematicians/ Saint-Venant.html 50. A.E. Jacobs: Phys. Rev. B 61, 658 (2000) 51. J.M. Ball, R.D. James: Arch. Ration. Mech. Anal. 100, 13 (1987) 52. R.D. James, Z. Zhang: this volume 53. Y. Wang, A. Khachaturyan: Acta Mater. 45, 749 (1997) 54. M.F. Thorpe, D.J. Jacobs, M.V. Chubynsky, J.C. Phillips: J. Non-Cryst. Sol. 266–269, 859 (2000); M. Micoulat, J.C. Phillips: Phys. Rev. B 67, 104204 (2003); D. Selvanathan, W.J. Bresser, P. Boolchand: Phys. Rev. B 61, 15061 (2000) 55. J. Zaanen, O. Gunnarsson: Phys. Rev. B 40, 7391 (1989); S.A. Kivelson, V.J. Emery: Physica C 235, 189 (1999); F. Kusmartsev: Europhys. Lett. 57, 567 (2000); D.I. Khomskii, K.I. Kugel: Europhys. Lett. 55, 208 (2001); M.B. Salomon, P. Lin, S.H. Chun: Phys. Rev. Lett. 88, 197203 (2002) 56. S.S. Saxena, P.B. Littlewood: Nature 119, 279 (2001); N. Mathur, P. Littlewood: Phys. Today (Jan.) 56, 25 (2003)

2 Spin, Charge, and Lattice Coupling in Multiferroic Materials

25

57. P. Welch, M. Muthukumar: Phys. Rev. Lett. 87, 218302 (2001); J. Schmalian, P.G. Wolynes: Phys. Rev. Lett. 85, 836 (2000) 58. A. Saxena, T. Cast´ an, A. Planes, M. Porta, Y. Kishi, T.A. Lograsso, D. Viehland, M. Wuttig, M. De Graef: Phys. Rev. Lett. 92, 197203 (2004) 59. A. Onuki: J. Phys. Soc. Jpn. 68, 5 (1999) 60. G.-M. Zhao, K.K. Singh, D.M. Morris: Phys. Rev. B 50, 4112 (1994); D. Zech, H. Keller, K. Conder, E. Kaldis, E. Liarokapis, N. Poulakis, K.A. M¨ uller: Nature 371, 681 (1994); D. Rubio-Temprano, J. Mesot, S. Jansen, K. Conder, A. Furrer, H. Mutka, K.A. M¨ uller: Phys. Rev. Lett. 84, 1990 (2000); G.-M. Zhao, K.A. M¨ uller: Nature 381, 676 (1996) 61. A. Lanzara, P.V. Bogdanov, X.J. Zhou, S.A. Kellar, D.L. Feng, E.D. Lu, T. Yoshida, H. Eisaki, A. Fujimori, K. Kishia, J.I. Shimayama, T. Hoda, S. Uchida, Z. Hussain, Z.X. Shen: Nature 412, 510 (2001) 62. M. Fiebig, K. Miyano, Y. Tomioka, Y. Tokura: Science 280, 1925 (1998); J. Gump, I. Finkler, H. Xia, R. Sooryakumar, W.J. Bresser, P. Boolchand: Phys. Rev. Lett. 92, 245501 (2004) 63. V. Hardy, A. Maignan, S. Herbert, C. Yaicle, C. Maitin, M. Hervieu, M.R. Lees, G. Rowlands, D.McK. Paul, B. Raveau: Phys. Rev. B 68, 220402 (2003); V. Podzorov, C.H. Chen, M.E. Gershenon, S.-W. Cheong: Europhys. Lett. 55, 411 (2001); A.K. Raychaudhuri: Proc. SPIE 5112, 144 (2003) 64. I.K. Bdikin, V.V. Shvartsman, A.L. Kholkin: Appl. Phys. Lett. 83, 4232 (2003) 65. D. Mihailovic, A.J. Heeger: Solid State Commun. 75, 319 (1990); D. Mihailovic, I. Poberaj, A. Mertelj: Phys. Rev. B 48, 16634 (1993) 66. S.R. Shenoy, V. Subrahmanyam, A.R. Bishop: Phys. Rev. Lett. 79, 4657 (1997) 67. E.L. Nagaev: Principles of Magnetic Semiconductors, 2nd edn. (Cambridge University Press, Cambridge, UK, 2000) 68. T. Lookman, S.R. Shenoy, G.V. Pai, A. Saxena: unpublished 69. S. Shimomura, N. Wakabayashi, H. Kuwahara, Y. Tokura: Phys. Rev. Lett. 83, 4389 (1999); I. Vasiliu-Doloc, S. Rosenkranz, R. Osborn, S.K. Sinha, J.W. Lynn, J. Mesot, O.H. Seeck, G. Preosti, A.J. Fedro, J.F. Mitchell: Phys. Rev. Lett. 83, 4393 (1999); Z. Islam, X. Liu, S.K. Sinha, J.C. Lang, S.C. Moss, D. Haskel, G. Srajer, P. Wochner, D.R. Lee, D.R. Haeffner, U. Welp: cond-mat/0312622; S.K. Sinha: Z. Kristall. 219, 143 (2004) 70. S.R. Shenoy, G.V. Pai, T. Lookman, A. Saxena: unpublished 71. T.V. Ramakrishnan, H.R. Krishnamurthy, S. Hassan, G.V. Pai: Phys. Rev. Lett. 92, 157203 (2004) 72. D. Heidarian, N. Trivedi: Phys. Rev. Lett. 93, 126401 (2004) 73. G. Kotliar, D. Vollhardt: Phys. Today (Mar.) 57, 53 (2004); M.L. Kulic: Phys. Rep. 338, 1 (2000)

3 Disorder in Magnetic and Structural Transitions: Pretransitional Phenomena and Kinetics T. Cast´an, E. Vives, L. Ma˜ nosa, and A. Planes, and A. Saxena

3.1 Introduction According to standard textbooks on thermodynamics [1], first-order phase transitions occur abruptly at given values of external control parameters such as temperature, pressure or applied field (stress, magnetic or electric). In real materials, however, these transitions rarely show such an ideal behaviour. Commonly, the expected sharp change of the order parameter is smoothed out and thus the transition spreads over a certain range of the external control parameter. Moreover, in a number of systems the transition is even preceded by anomalies, typically detected in the response to certain excitations, which may arise from local symmetry breaking perturbations. Actually, all the anomalies which originate from the tendency of the system to perform anticipatory visits to the low-symmetry phase are related to the so-called pretransitional effects which can be defined as phenomena that announce that the system is preparing for a phase transition before it actually occurs. In what follows, transitions exhibiting the behaviour described above will be denoted extended first-order transitions. Nowadays, it is accepted that the influence of disorder in the form of randomly quenched defects or impurities, almost always present in any solid system, is at the origin of all these deviations with respect to the ideal first-order behaviour. In this sense, there is an emerging point of view in materials physics that disorder, understood in a broad sense (as imperfections, inhomogeneities, point defects, boundaries, etc.), is intrinsic to real materials. In fact, a wide variety of interesting properties in materials have only been understood when considering disorder as something inherent to the system [2]. There are two essential issues that enable us to account for the above-mentioned deviations from the ideal first-order transitions. On one side, the system must sensitively respond to quenched defects by, for instance, giving rise to variations of the effective local transition temperature. On the other side, as already pointed out by Imry and Wortis [3], when the transition starts the growth of the correlations is blocked by defects. More recently, the existence of an associated

28

T. Cast´ an et al.

disorder induced transition related to the change from a sharp to a smooth transition has been predicted [4]. It is worth noting that in all cases long-range interactions, as arising from surface (or interface) effects through compatibility conditions [5, 6], play a fundamental role. They enable a cooperative response of the system to occur in the pretransitional regime by connecting the different perturbed regions. Moreover, they provide the necessary conditions for the transition kinetics to be athermal as opposed to thermally activated [7]. In the present chapter we will discuss and try to establish the general requirements for magnetic and structural materials to show pretransitional effects. Furthermore, the effect of disorder on the transition kinetics will also be analysed. The observed phenomenology will be illustrated with examples for magnetic and non-magnetic materials undergoing martensitic transitions (MTs). In addition, the possibility of simultaneous structural and magnetic precursors will be shown in the case of materials displaying magnetostructural transitions.

3.2 Disorder Distribution As mentioned earlier, the point of view adopted in this work is that disorder is intrinsic to real materials. Generically, one might characterize the disorder in terms of a distribution such as sketched in Fig. 3.1. Intuitively the effects of the tails can be separated from the center of the disorder distribution as follows: The typical disorder affects the central values and fluctuations of most of the bulk properties, and will be important in determining the properties and kinetics of the transition. The tails of the distribution include non-typical disorder in the sense of non-dominant contributions such as from boundaries and extreme values, and become more relevant for small systems. It is expected that the properties affected by this kind of disorder will be much more sample dependent than in the case of typical disorder. We suggest that pretransitional effects are mainly determined by non-typical disorder. Based on this viewpoint, in this chapter we separately discuss the transition kinetics and pretransitional effects.

Fig. 3.1. Schematic representation of the disorder distribution in materials. N is the size of the system

3 Disorder in Magnetic and Structural Transitions

29

3.2.1 Pretransitional Phenomena Pretransitional effects have been observed in a wide variety of materials. Typical examples include elastic constant and phonon anomalies in High Tc superconductors [8], ferroelectrics [9], manganites [10, 11] and shape memory alloys [12]. In addition, many of these anomalies are associated with precursors such as satellites in diffraction patterns and modulated structures [13–17]. It must be noted that these structures cannot be understood as a coexistence of different phases but rather as a mesoscopic texture within a single thermodynamic phase because variations in the physical properties occur over very small distances (at the nanometer scale) [18]. Here, we shall focus on solids (magnetic and non-magnetic) undergoing MTs and we begin by clarifying what we mean by pretransitional phenomena. The high temperature phase of most materials undergoing an MT exhibits a cubic bcc-based open structure. This turns out to be of crucial importance in the pretransitional behaviour of the material already well above the (martensitic) transition temperature (TM ). Among pretransitional effects, we shall call precursors those that involve structural changes – either dynamic or/and static – and consequently might influence the way in which the system approaches the incoming MT. To describe pretransitional effects that do not involve structural changes we shall use the term premonitory. These originate from the existence of low restoring forces along specific directions related to the symmetry of the high temperature cubic phase. More precisely, in bcc systems to the low resistance against distortions of the {110} planes along the 1¯ 10 directions which is a signature of the intrinsic instability of the lattice already pointed out by Zener [19] and Friedel [20]. Such a low resistance is reflected by the low energy of the corresponding transverse acoustic phonon branch TA2 and the low value of the shear elastic constant C
= (C11 −C12 )/2. It naturally follows that premonitory effects are inherent to the cubic structure rather than to the MT itself. They are even observed in materials that do not transform martensitically [16]. Nevertheless, they are at the origin of a wide variety of observed precursors and in the case of martensitic materials they participate actively in the structural instability. 3.2.2 Premonitory Behaviour: Temperature Softening The response of a solid to long wavelength vibrational excitations is given by the elastic constants. A quite general characteristic of bcc materials is the low value of the elastic constant C
compared with other elastic moduli. Often, this is accompanied by a low value of the energy of phonons propagating along the [110] direction with [1¯ 10] polarization (i.e. the TA2 phonon branch) over the entire Brillouin zone. In effect, these two features are a consequence of the open character of the bcc structure which enables easy motion of the {110} planes. This incipient mechanical instability provides an easy channel for the MT. On the other hand, the low energy of the overall TA2 phonon branch

30

T. Cast´ an et al.

Fig. 3.2. (a) Temperature softening of the elastic constant C  in Ni63 Al37 and Cu2.742 Al1.105 Ni0.152 . (b) Corresponding behaviour of the elastic anisotropy A. Data from [23, 24]

results in a high vibrational entropy which, from a thermodynamical point of view, is responsible for the stability of the bcc phase [12, 21]. Regarding MT the relevant aspect is the softening of both C
and the TA2 phonon branch with cooling. In addition, there appears to be an intimate relationship between the value of the elastic anisotropy A = C44 /C
and the structural instability in the sense that the value of A at TM has some degree of universality [12,22]. Figure 3.2 illustrates the temperature softening of C
(a) for Cu–Al–Ni and Ni–Al crystals and the corresponding increase in the elastic anisotropy A and (b) when approaching the transition TM from above. Eventually, certain martensitic materials might exhibit a dip on the TA2 phonon branch. That is, the phonon branch shows a minimum at a particular value of the wave vector q = 0. This dip is particularly noticeable when the energy of the whole branch is not very low (although still lower than the other branches). Examples are presented in Fig. 3.3 for Ni–Al, Cu–Al–Be and NiTi(Fe) crystals. In that case, the dip becomes more pronounced when cooling which means that the anomalous phonon(s) soften more than the rest. The origin of this phonon dip anomaly is not intrinsic to the cubic symmetry but is due to the particular shape of the Fermi surface which has nesting features at specific wave vectors [25]. It is worth noting that for alloys, the position of the dip depends on composition [26]. Since at TM there is no complete softening of any of the phonons (including the elastic limit q = 0), it has been suggested that the transition is due to a highly anharmonic coupling between both homogeneous (q = 0) and inhomogeneous (q = 0) distortions [28]. It is worth remarking that the presence of a dip on the TA2 branch is not a necessary condition for the subsequent MT. Nevertheless, the existence of low energy excitations in the system should affect the actual transformation

3 Disorder in Magnetic and Structural Transitions

31

Fig. 3.3. TA2 phonon branch in Ni0.625 Al0.375 , Cu0.7373 Al0.2272 Be0.0355 and NiTi (Fe). Right panel shows the temperature softening of the phonon dip anomaly for NiTi(Fe). Lines are guides to the eyes. Data from [26, 27]

path. In this sense it has been suggested [26] that the wave vector of the dip anomaly is related to the periodicity of the close-packed planes in the martensitic structure. This is, however, an open issue since there are materials for which such stacking periodicity does not seem to be related to the dip anomaly (if there is any significant softening) [12]. 3.2.3 Structural Precursors The signature of the softening behaviour described in Sect. 3.2.2 is seen in the form of diffuse scattering. Actually, the low value of C
will produce strong ridges of thermal diffuse scattering running in the [110] directions [29]. Anisotropic thermal scattering has been observed in electron [17, 30–34], X-ray [35] and neutron [26] diffraction experiments. As an example, we show in Fig. 3.4 the electron diffraction (ED) patterns obtained for Ni–Al and Cu–Zn–Al crystals [33]. It is worth noting that Cu–Zn–Al, which has larger elastic anisotropy (A) [12], exhibits stronger anisotropic diffuse scattering than Ni–Al. In agreement with the ideas exposed above, diffuse scattering is observed in a wide variety of cubic materials with large values of A, independently of whether or not they exhibit a MT at finite temperatures. In some cases, additionally to the thermal displacement fluctuations mentioned above, there exist static or long-lived fluctuations. As will be stated later, these fluctuations originate from the presence of quenched-in disorder (imperfections, inhomogeneities, etc.) that pin the fluctuations and are at the origin of the so-called tweed precursor strain modulations or simply structural tweed. It is the natural long-range elastic response to local disorder in cubic systems with large elastic anisotropy. Below we shall generalize this concept to other systems with modulations in other physical variables.

32

T. Cast´ an et al.

Fig. 3.4. ED pattern of samples with [001] normal of Ni–Al (left) and Cu–Zn–Al (right). Data from [33]

Fig. 3.5. HTEM strain contrast of samples with [001] normal for Ni–Al (left) and Cu–Zn–Al (right). Data from [33]

The contribution of such static (or long-lived) displacement fluctuations to the diffuse streaking shows up in real space strain contrast TEM images [16,30, 33,36,37]. It consists of a dense array of quasi-periodic striations lying parallel to {110} plane traces giving rise to the tweed pattern which is cross-hatched with correlations along the diagonals. The tweed striations are perpendicular to the [110] diffuse streaking and have been identified to correspond to strain waves of [110]1¯ 10 displacements [16]. It is very interesting to compare the characteristics of tweed in Ni–Al with that observed in Cu–Zn–Al, both shown in Fig. 3.5. We remind that whereas Ni–Al shows a significant dip accompanied with satellites in diffraction experiments, Cu–Zn–Al does not show any significant phonon anomaly but the whole TA2 branch is very low lying in energy [38]. In both the materials the elastic anisotropy is quite large and the characteristics of the underlying modulation are the same. Nevertheless, whereas in Ni–Al tweed is mostly static, in Cu–Zn–Al the dynamic component is predominant [33]. We suggest this difference is due to a local coupling between the strain and the anomalous phonon in the case of Ni–Al. For those materials exhibiting a significant dip anomaly one observes elastic scattering at the same wave vector as the phonon softening. This is the central peak reported in neutron scattering experiments and corresponds to the satellites observed in either X-ray [39], or electron [33, 40, 41] or neutron [42]

3 Disorder in Magnetic and Structural Transitions

33

diffraction patterns. These satellites are broad peaks which indicate that they come from correlations which do not attain true long-range order. Nevertheless, there is an underlying structural modulation, either static or dynamic, whose wavelength is predicted by the dip anomaly and that eventually might correspond to the stacking fault periodicity of the martensitic phase. 3.2.4 Coupling to Magnetism Recently, magnetic shape memory alloys have received an increasingly high interest because of the possibility of inducing large deformations by application of an external magnetic field. This effect was first discovered in Ni–Mn– Ga [43]. The corresponding phase diagram is shown in Fig. 3.6 as a function of the electron concentration per atom (e/a). An extensive amount of research has been devoted to finding alternative materials showing magnetic shape memory effect. Among others some examples include Co–Ni–Al, Co–Ni–Ga, Ni–Mn–Al, Fe–Mn–Ga, etc. The pretransitional behaviour of these materials shows specific features which are due to coupling between structural and magnetic degrees of freedom. Although, up to now, experimental data of vibrational properties are only available for Ni2 MnGa, these bcc-based alloys should conform to the general scenario described above: The temperature softening of the TA2 phonon branch at q = 0 and eventually a dip at q = 0. In the case of Ni–Mn–Ga, a dip anomaly is observed at a value of q which depends on composition [44, 45] as illustrated in Fig. 3.7 for e/a = 7.56 (left panel) and e/a = 7.71 (right panel) crystals. Associated with the dip anomaly, structural precursors such as broad satellite peaks are observed. The signature of the coupling to magnetism is primarily observed as an enhancement of the softening of the anomalous phonon (dip)

Fig. 3.6. Phase diagram of Ni–Mn–Ga as a function of the electron concentration per atom (e/a). P stands for paramagnetic, F, ferromagnetic and M, for martensitic. The dotted line denotes the intermediate transition (TI ), TM is the martensitic transition and TC is the Curie line

34

T. Cast´ an et al.

Fig. 3.7. Phonon dip anomaly for Ni52.0 Mn23.0 Ga25.0 (left) and Ni50.5 Mn29.5 Ga20.0 (right) crystals, which transform, respectively, to different modulated structures. Note the difference in the location of the dip. Lines are guides to the eye. Data from [44]

below the Curie point (TC ) [44,46]. This is shown in Fig. 3.8 for the same samples as before (left panel of Fig. 3.7). Concerning the long wavelength limit (Fig. 3.8b), apparently, there is no change in the temperature dependence of C
when crossing the Curie point (TC ) [47]. It has been suggested that, in the absence of external magnetic field, the magnetic domain structure below TC results in an almost zero magnetization and therefore no significant change in the macroscopic elastic response is expected [47]. The magnetoelastic coupling is revealed in this case by an increase of the elastic constants with applied magnetic field [48]. We now focus on a new feature, that is the upturn observed in both the anomalous phonon and in C
at a temperature (TI ) slightly above the MT for the low e/a crystal. This is the signature of a phonon condensation which gives rise to the development of a modulated phase, which in turn is evidenced by the fact that the broad satellites become true Bragg peaks [49]. Such an intermediate phase consists of a micromodulated domain structure without resulting in a macroscopic deformation so that the cubic symmetry is preserved on average. It has been shown that magnetoelastic coupling is responsible for the occurence of such intermediate phase [50] and that the associated phase transition (TI ) is of first order [51, 52]. This behaviour is characteristic of multi-stage structural transformations [53]. For the large e/a crystal the phonon condensation (and therefore the intermediate phase) is suppressed by the occurrence of a structural instability which leads (directly) to the low temperature martensitic phase.

3 Disorder in Magnetic and Structural Transitions

35

Fig. 3.8. (a) Energy of the anomalous phonon (dip) as a function of temperature for two Ni–Mn–Ga crystals. Lines are guides to the eye. Note the change in slope at TC for the two cases and the upturn for the sample (low e/a) with the intermediate phase transition at TI . Data from [44]. (b) Relative change of shear elastic constant C  as a function of temperature for the same two Ni–Mn–Ga crystals. Data from [47]. Again, note the upturn at TI in C  for the low e/a crystal

Note the remarkable kink observed in the softening of the anomalous phonon at TC (see Fig. 3.8a) for the sample with high value of e/a. From Fig. 3.6, it follows that this intermediate modulated phase exists only for low e/a, that is in the region where the MT is far below the Curie point. 3.2.5 Ferroic Precursors: Magnetic and Others In this section we generalize the tweed concept to other systems than purely structural, and special attention is given to magnetic systems. The starting point is a general framework aimed to understand precursor modulations in the so-called ferroic materials [54], that is, modulations in either strain, or magnetization or polarization observed above the ferroic transition. We remark that since such modulations occur prior to a phase transition one expects the properties of the incoming phase to be affected by the actual initial state of the system. Recent advances in high resolution imaging of magnetic domain patterns, such as magnetic force microscopy [55] and Lorentz transmission electron microscopy (LTEM) [56], have revealed fascinating modulated magnetic patterns both above and below the Curie temperature in certain magnetic alloys. In contrast to the structural tweed discussed above, the magnetic modulations that occur as precursors to the magnetic transition give rise to a stripe-like pattern. Figure 3.9 shows the magnetic pattern observed by Lorentz microscopy in Co–Ni–Al above the Curie point. Very recently [58], TEM observations in Co2 NiGa alloys showing both strain and magnetic modulations have been presented, and a model that explains these observations has been proposed. From the observation of the

36

T. Cast´ an et al.

Fig. 3.9. Magnetic tweed-like modulations observed by Lorentz microscopy in Co0.380 Ni0.330 Al0.290 above the Curie point. Data from [57]

magnetic modulations it naturally follows that the tweed concept is not just structural but applicable to a much broader class of materials. In addition, it is shown that, independently of specific details of the pattern or the physical variable involved in modulation, the origin of tweed lies in very general requirements, likely to be fulfiled in quite different systems undergoing phase transitions. For instance, polar (or dielectric) tweed has been observed in ferroelectrics [15, 59]. Extending ideas put forth earlier by other authors in the context of (purely) structural tweed [5, 60], we suggest that the tweedlike modulations above the transition (structural, magnetic or other) are a natural cooperative response in systems that are sensitive (in the sense of, e.g. phonon softening, “susceptibility” or other response functions) to local symmetry breaking perturbations (e.g. due to statistical disorder) assisted by anisotropic long-range interactions. The long-range nature of such interactions (elastic, magnetic or other) connects the different perturbed regions while the anisotropy determines the specific modulations of the resulting pattern. We note that statistical compositional disorder is intrinsic to alloys and therefore they are the most probable candidates to exhibit such phenomena. The magnetic tweed-like modulations could conceivably arise in two different ways: (1) Purely due to magnetic disorder without any structural degrees of freedom being involved. This can be called purely magnetic tweed or simply magnetic tweed . It has been observed [57] above the Curie temperature (TC ) in Co0.38 Ni0.33 Al0.29 alloys, see Fig. 3.9. (2) Due to structural or compositional disorder which produces a structural tweed above the structural transition temperature (TM ) in magnetic alloys undergoing a MT. A coupling of strain with magnetism can then lead to a magnetic modulation both above and/or below TC . This kind of tweed is termed magnetoelastic tweed . It has been observed by LTEM in ferromagnetic (Heusler) Ni2 MnGa [61, 62] and Co2 NiGa alloys [63]. By using two-dimensional models on a square lattice it can be shown [58] that magnetic tweed patterns are stripe-like whereas magnetoelastic tweed produces cross-hatched patterns.

3 Disorder in Magnetic and Structural Transitions

37

The concept of a structural or a magnetic tweed can be generalized to a “ferroelastic tweed”. A “proper” ferroelastic tweed would correspond to strictly the structural tweed [5, 6]. However, magnetic tweed (in magnetic or magnetoelastic materials) and polar tweed in ferroelectrics [15, 59, 64] would correspond to an “improper” ferroelastic tweed because both the disorder and the long-range interaction may arise from a physical variable other than strain, e.g. magnetization or polarization. Improper ferroelastic tweed may also be observable in Jahn–Teller distorted colossal magnetoresistance (CMR) materials, giant magnetoresistance compounds [65], high Tc superconductors [66] and other perovskites with disorder in, for instance, octahedral tilts. The origin of the anisotropic long-range interaction in proper ferroelastic (or structural) tweed is the elastic compatibility constraint [5, 6] whereas polar or magnetic dipole interaction can arise from the crystal surface.

3.3 Transition Kinetics In the previous sections, we have dealt with the effect of the non-typical disorder in materials undergoing extended first-order phase transitions. We now turn our attention to the effect of typical disorder on these transitions. The most evident effect is that, instead of occurring at a given point of the phase diagram, they spread in a certain region of the control parameters. In this region there is a coexistence of the high and low temperature phases. Usually these driving parameters are the temperature or conjugated fields of the order parameters. In practice, distinction between pretransitional effects and the actual transition process is not always evident. We propose, as a general criterion, that significant hysteresis occurs in the transition region, whereas pretransitional effects are almost reversible. 3.3.1 Athermal Transitions The transitions to be considered are always associated with the existence of high energy barriers, which separate the parent and product phases that must be overcome for the transition to proceed. Figure 3.10 shows schematically a phase diagram for such systems as a function of temperature T and a generic driving field H. We indicate the region of pretransitional effects and the hysteresis region, that typically has a width ∆H (or ∆T ). Indeed, in the irreversible region the actual state of the system is metastable and history dependent. It is also worth noting that, although we have plotted a single axis for the field, in multiferroic materials (such as those with two or more coupled ferroic properties, e.g. magnetoelastic materials) there can be more than one different external fields coupled to the relevant order parameters. The following main characteristics of such extended first-order phase transitions must be emphasized:

38

T. Cast´ an et al.

Fig. 3.10. Schematic phase diagram of an extended first-order phase transition as a function of the conjugated field and temperature. Pretransional and transitional regions are indicated. Arrows denote that the transition can be induced either by cooling or by applying an external field

• The existence of long-range forces of elastic, magnetic or electric nature arising from compatibility conditions (or dipolar forces) associated with the system boundaries and/or interfaces. The external force driving the transition must then compete with the whole system and not only with a small microscopic domain. Thus thermal fluctuations play, in general, only a minor role, which means that the time scale of the thermal activation processes τth is very large. This time scale is proportional to exp[−Eb /kB T ], where Eb is a measure of the free-energy barriers. • The existence of quenched disorder. In the scenario described in the previous paragraph it is clear that any perturbation that locally modifies the profile of the energy landscape will be crucial for the transitions to proceed. Therefore, as a consequence of quenched disorder the energy barriers are inhomogeneously distributed in the system. This gives rise to the existence of a complex energy landscape with multiple energy minima where almost divergent barriers may coexist with low energy paths. Thus, in this class of transitions quenched disorder of any kind (dislocations, vacancies, composition fluctuations, impurities, etc.) is expected to strongly modify the transition path and the hysteresis features. In many cases disorder cannot be considered strictly quenched, since their characteristics (amount of disorder and spatial distribution) can be modified by cycling through the transition. In this case modification of disorder enables to establish an optimal transition path that avoids high energy barriers which arise from long-range forces. Moreover, in some cases aging effects have been reported (for instance, associated with reordering processes [67]), but on time scales of τth . • As a consequence of these complex free-energy landscapes, the intermediate and the final state of the transition exhibit a mesoscopic domain pattern which in some cases displays self-similar properties and a lack

3 Disorder in Magnetic and Structural Transitions

39

Fig. 3.11. Optical micrograph with polarized light on the surface of a Cu–Al–Zn sample after cooling through MT. Domains corresponding to different martensite variants with many different characteristic lengths coexist

of characteristic scales. An example is presented in Fig. 3.11, that corresponds to the final structure of a MT in a Cu–Zn–Al single crystal. Note the multiple lengthscales of the martensitic domain structure. • The response of the system to the driving parameter shows, in a broad range of driving rates, avalanche behaviour [68]. Different experimental detection techniques reveal that the evolution of the transition is characterized by burst-like events separated by inactivity periods. Such events correspond to jumps from one metastable state to another involving dissipation of energy. In many cases the properties of the avalanches reveal a lack of characteristic scales within a broad range of time, sizes and energies. Although this makes it difficult to properly define a characteristic time scale associated with the avalanches, the duration of the longest detected avalanche event τav is typically much smaller than the time scales associated with thermal activation processes τth . • The last common feature is the dependence of the hysteresis properties ˙ A convenient way to and the avalanche behaviour on the driving rate H. describe the dependences with the driving rate is by defining a third time ˙ In most cases, when τdr is much larger scale τdr which is given by ∆H/H. than τav , the avalanche response of the system is essentially deterministic and reproducible (rate independent hysteresis), but in other cases when τdr and τav compete, rate dependent hysteresis effects are observed. These factors may simultaneously exist in MT [69, 70] and in some ferromagnetic systems [71] at low enough temperatures. Interestingly, similar behaviour is also observed in perovskite manganites [72–74]. The dynamics of such transitions have been termed as athermal . In practice, this means that the transition driven by a temperature change shows similar kinetic effects as compared to those observed in the transition driven by an external field (applied stress, magnetic field, etc.). These two paths are indicated in Fig. 3.10.

40

T. Cast´ an et al.

Note that the case of a simple uniaxial ferromagnetic material is a particular case in which the phase transition region in Fig. 3.10 is horizontal and symmetric with respect to the temperature axis. Thus the athermal transition can only driven by external (magnetic) field. 3.3.2 Modelling Athermal Transitions From a more fundamental point of view, athermal represents a limiting case in which the time scale of thermal activation τth is much larger than τav and τdr . A first attempt to analyse this situation has been recently considered by P´erez-Reche et al. [7, 70]. The proposed model enables a good understanding of the possible interplay of the three time scales introduced above. However, since the existence of spatial inhomogeneities which arise from disorder are not explicitly included in the model, it only allows for the description of the onset of the transition but it is not appropriate for the study of its subsequent evolution. Therefore, one of the main drawbacks of the model is that it does not provide an explanation for the avalanche behaviour. The approach is based on a mean-first-passage time model [75] which assumes that the energy barrier (associated with the first transforming domain) changes monotonically with the driving field H. The barrier Eb (H) is infinite for H > Hh and vanishes for H ≤ Hl . Only within this interval Hh −Hl thermal fluctuations can play a role in determining the onset of the transition. Therefore, the transition onset field Ht (or transition temperature) is a random variable that must be averaged over different equivalent experimental realizations. From the dependence of the average transition field and its standard deviation on the driving field rate, the fields Hh and Hl can be estimated. Within the preceding framework, the ideal athermal case corresponds to the limit in which the interval Hh − Hl → 0 not only for the first transforming domain but also for all the subsequent barriers to be overcome. This is equivalent to the assumption that τth → ∞. In this case the detailed response of the order parameter m(H) is time independent and its derivative must satisfy: dm ˙ dm = H. (3.1) dt dH Thus, although dm/dt may exhibit a dependence on the driving rate, the product H˙ −1 dm/dt must be rate independent. This scaling behaviour can be used to test the validity of the athermal hypothesis of a given transition. Figure 3.12 shows an example of the scaling behaviour corresponding to a first order MT in a Cu–Zn–Al alloy. The dynamics is monitored by detecting the acoustic emission (AE) generated during the transition. 3.3.3 Avalanche Dynamics As mentioned above, avalanches are typically detected in a broad range of energy, size and time scales. This supported the idea of the existence of criticality in such systems. Experiments for the detection of avalanches are usually

3 Disorder in Magnetic and Structural Transitions

41

Fig. 3.12. (a) Number of AE signals per unit time (ν) as a function of temperature recorded during continuous cooling runs for a Cu–Zn–Al sample at different rates from T˙ = 0.5 K/min−1 (upper curve) to 5 K/min−1 (lower curve). (b) Scaled AE as a function of temperature. Inset shows an enlarged view of the tail region

performed on systems which exhibit a strong athermal character. In this case, avalanches only occur when the system is driven by the control field. If the field is kept constant no response is observed. However, in some cases avalanches have been detected at constant field and temperature as a function of time. In these cases, the systems are sufficiently close to the athermal limit so that avalanches can still be discriminated by an experimental threshold, although thermal fluctuations cannot be ruled out. An example is presented in Fig. 3.13 (isothermal transformation in Cu–Al–Ni). This discussion suggests that, in practice, the athermal character of a system depends not only on its intrinsic features but also on the resolution of the experimental technique being used and the range of driving rates that are experimentally accessible. The experimental analysis of the avalanche properties has been performed by means of several techniques which enable its study at different scales. These include AE for the detection of the elastic waves associated with the avalanches in MT, Barkhausen noise recordings in magnetic systems, calorimetry, or even direct imaging methods that allow an analysis of the spatial scales of the avalanches. The statistical distributions of avalanche size, duration, energy, etc. are commonly described in terms of the following probability distribution with two free parameters,

42

T. Cast´ an et al.

Fig. 3.13. Number of AE signals per unit time (ν) as a function of time for MT of a Cu–Al–Ni sample recorded during stepwise cooling. Inset shows a detailed view of AE recorded isothermally

e−λZ Z −x , p(Z) =  Zmax −λZ Z −x dZ e Zmin

(3.2)

where Zmin and Zmax are the minimum and maximum experimentally detectable values of the magnitude Z, x is a characteristic power-law exponent and λ determines the exponential correction to the power-law behav−1 , the exponential correction can iour. When |λ| is much smaller than Zmax be neglected and the system is assumed to exhibit critical behaviour. When this is not the case, for λ > 0 the system is called subcritical while for λ < 0 it is called supercritical. Figure 3.14 shows an example of the distribution of the amplitudes of AE avalanches during the MT in a Ni–Mn–Ga alloy. It has been shown that the distribution of avalanches in structural [76] and magnetostructural [77] transitions exhibits a dependence on cycling. The first transition after proper annealing of the sample at high enough temperature usually exhibits a clearly non-critical distribution of avalanches (supercritical or subcritical). The effect of a subsequent cycling is to modify the value of |λ| until a stable critical distribution is reached with well-defined values of the exponents. In MT this has been related to the increase of the density of dislocations which in turn results in an increase of favourable nucleation centers for the transition. For Barkhausen noise in magnetic systems and AE in MT, different universality classes have been proposed that are characterized by sets of critical exponents. Nevertheless, dependences on temperature [78] and driving rate [79, 80] have also been reported. Thus, whether the exponents are really universal or not is still a question under debate.

3 Disorder in Magnetic and Structural Transitions

43

Fig. 3.14. Distribution of AE avalanches during MT in a Ni–Mn–Ga alloy. Lines are fits to the distribution (3.2)

3.3.4 Modelling Avalanches The prototypical model for understanding avalanche behaviour in first-order athermal phase transitions is the Random Field Ising Model with T = 0 metastable dynamics [4]. This is a lattice model with spin variables Si (which can represent domains or small regions of the material) localized on each lattice site and coupled ferromagnetically. Quenched Gaussian random fields with zero mean and standard deviation σ are defined on each site. They represent the effect of compositional and morphological disorder. The model is subjected to an external field. When the system is driven by very high values of the field H, the saturated state with all Si = +1 evolves following a deterministic dynamics consisting of a local energy relaxation of single spins. Once a spin flips, this may trigger the flip of neighbouring spins and initiate an avalanche. During avalanches the driving field is kept constant until a new metastable situation is reached. This kinetics, which has also been called adiabatic, conveniently mimics athermal dynamics in real systems and corresponds to a situation of rate independent hysteresis in which τav is much smaller than τdr and τth . By numerical simulations one obtains hysteresis loops consisting of a sequence of avalanches joining metastable states. These avalanches are in general microscopic which means that in the thermodynamic limit, the associated jumps involve the transformation of a vanishingly small fraction of the system. Interestingly, the loops exhibit a transition from being smooth when σ > σc to showing a ferromagnetic macroscopic discontinuity when σ < σc . At σc the distribution of avalanches is critical and can be characterized by a well defined exponent. Other similar models including random bonds, random anisotropy, vacancies, etc. have also been studied. They all display similar behaviour and it

44

T. Cast´ an et al.

has been shown that the critical exponents exhibit a high degree of universality [81]. These models reveal that, indeed, the existence of disorder is one of the key ingredients in the scenario described in the previous sections. Nevertheless, we note that they do not take into account the existence of long-range interactions arising from compatibility conditions which are essential to understand the formation of multidomain structures. Moreover, one may speculate that longrange forces are also responsible for the observed learning process during cycling in which the system seeks an optimal path connecting the high and low-symmetry phases which avoid high energy barriers. After the learning period, a stationary state for which the transition is reproducible from cycle to cycle should be reached. This state should be determined by a compromise between the increase of energy associated with the increase of disorder and the corresponding reduction of metastability (which supposes a reduction of dissipated energy). The tendency of the distribution of avalanches to become critical after a certain number of cycles supports that learning is induced by long-range interactions. Recently, a more realistic approach to modelling avalanches in MT has been undertaken by Ananthakrishna and collaborators [82]. They proposed a model based on a free energy functional which includes an effective long-range term that describes the transformation induced strain–strain interaction and an inhomogeneous stress field (Gaussian distributed) adequate to mimic the nucleation of martensitic domains in the vicinity of defect sites. Dissipation is accounted for by means of a Rayleigh dissipative functional. While elastic compatibility is, for the sake of simplicity, not properly taken into account, the model has the ability of reproducing most of the dynamical signatures typical of MT in shape memory alloys including the athermal character. The model generates dynamical disorder during the transformation which occurs through energy bursts which in turn are identified with AE signals. These bursts are correlated with the growth and shrinkage of martensitic plates. In agreement with the experiments, they obtained that size and duration of AE bursts follow power-law statistics with exponents comparable to those reported for Cu-based alloys. An interesting result of the model relates the evolution of the martensitic morphology with thermal cycling during a certain training period [83]. After this initial number of cycles the model exhibits repetitive bursts of energy under successive cycles. This cycling evolution, which is in agreement with cycling experiments [76], is shown to be mediated through strain–strain long-range interactions. The authors concluded that during the training period the long-range term has a tendency to smooth out higher energy barriers in the free energy landscape which induces a transformation pathway along a unique set of low energy metastable configurations.

3 Disorder in Magnetic and Structural Transitions

45

3.4 Conclusion A vast majority of functional materials invariably have some form of disorder which crucially affects their properties. In particular, disorder extends the region of the first-order transition. Disorder typically has a distribution sketched in Fig. 3.1. The central part of the distribution impacts the kinetics of the transition such as the athermal nature and avalanche behaviour of AE in MT. In contrast, the tails of the distribution are responsible for pretransitional effects, including prescursor modulations, taking advantage of the premonitory softening behaviour. We illustrated these points in both purely elastic shape memory alloys (Ni–Al, NiTi, Cu–Al–Ni, Cu–Zn–Al and Cu–Al–Be) and magnetoelastic alloys (Ni–Mn–Ga, Co–Ni–Ga). The coupling of structure to magnetism enhances the softening of the anomalous phonon below the Curie point and may also lead to a magnetoelastic tweed. In some alloys (e.g. Co–Ni–Al) a purely magnetic modulation (i.e. magnetic tweed) can be present above the Curie temperature. Thus, the notion of tweed appears to be very generic and the tweed can exist in such diverse materials as ferroelectrics, CMR manganites and high-temperature superconductors. We pointed out the common ingredients in these materials that are responsible for tweed. These materials are sensitive to local symmetry breaking perturbations (i.e. disorder). Therefore, various tweed patterns are stabilized by anisotropic long-range interactions which may arise from elastic compatibility constraints and from the surface of the crystal. We emphasized the athermal nature of the transition in some of these materials and discussed the main characteristics including complex free energy landscapes due to the presence of quenched disorder and the dependence of hysteresis as well as avalanche behaviour on the driving rate (of an external field). We also outlined a simple model for understanding the avalanches. This chapter represents an attempt to unify some disorder related concepts in ferroic materials but clearly many questions remain open in this field. Acknowledgements We are grateful to the members of our group, Xavier Illa, Jordi Marcos, Francisco-Jos´e P´erez-Reche, Marcel Porta and Carlos Triguero, who actively participated in obtaining some of the results presented in this chapter. We also thank our colleagues M. Acet (Duisburg University), M. Morin (INSA – Lyon), M. Stipcich (IFIMAT – Tandil) and C. Stassis and J. Zarestky (Ames Laboratory) for valuable discussions. This work was supported by Spanish CICYT (MAT2001-3251), Catalan DURSI (2001SGR00066) and in part by the US Department of Energy.

46

T. Cast´ an et al.

References 1. H.B. Callen: Thermodynamics and an Introduction to Thermostatics, 2nd edn. (John Wiley & Sons, New York, 1985), pp. 215–244 2. A.J. Millis: Solid State Commun. 126, 3 (2003) 3. J. Imry, M. Wortis: Phys. Rev. B 19, 3580 (1979) 4. J.P. Sethna, K. Dahmen, S. Kartha, J.A. Krumhansl, B.W. Roberts, J.D. Shore: Phys. Rev. Lett. 70, 3347 (1993) 5. S. Kartha, J.A. Krumhansl, J.P. Sethna, L.K. Wickham: Phys. Rev. B 52, 803 (1995); and references therein 6. S.R. Shenoy, T. Lookman, A. Saxena, A.R. Bishop: Phys. Rev. B 60, R12537 (1999); T. Lookman, S.R. Shenoy, K.Ø. Rasmussen, A. Saxena, A.R. Bishop: Phys. Rev. B 67, 024114 (2003) 7. F.J. P´erez-Reche, E. Vives, L. Ma˜ nosa, A. Planes: Phys. Rev. Lett. 87, 195701 (2001) 8. R.J. McQueeney, Y. Petrov, T. Egami, G. Shirane, Y. Endoh: Phys. Rev. Lett. 82, 628 (1999) 9. G. Shirane, Y. Yamada: Phys. Rev. 177, 858 (1969) 10. D. Louca, T. Egami, E.L. Brosha, H. R¨ oder, A.R. Bishop: Phys. Rev. B 56, R8475 (1997) 11. W. Reichardt, M. Braden: Physica B 263–264, 416 (1999) 12. A. Planes, L. Ma˜ nosa: Solid State Phys. 88, 159 (2001); and references therein 13. Z. Jincang, L. Lihua, D. Cheng, L. Jianqi, C. Hong, L. Xigui, C. Guosheng: Phys. Rev. B 65, 54513 (2002) 14. J.M. Tranquada, B.J. Sternlieb, J.D. Axe, Y. Nakamura, S. Uchida: Nature 375, 561 (1995) 15. X. Meng, K.Z. Baba-Kishi, G.K.H. Pang, H.L. Chang, C.L. Choy, H.S. Luo: Phil. Mag. Lett. 84, 191 (2004) 16. L. Tanner: Phil. Mag. 14, 111 (1966) 17. S.M. Shapiro, J.Z. Larese, Y. Noda, S.C. Moss, L.E. Tanner: Phys. Rev. Lett. 57, 3199 (1986) 18. N. Mathur, P. Littlewood: Nature Mater. 3, 207 (2004) 19. C. Zener: Phys. Rev. 71, 846 (1947) 20. J. Friedel: J. Phys. (Paris) Lett. 35, L59 (1974) 21. W. Petry: J. Phys. IV 5, C2 (1995) 22. G. Mazzolai, A. Biscarini, B. Coluzzi, F.M. Mazzolai, A.R. Ross, T.A. Lograsso: Mater. Sci. Eng. A 370, 497 (2004) 23. L. Ma˜ nosa, M. Jurado, A. Planes, J. Zarestky, T. Lograsso, C. Stassis: Phys. Rev. B 49, 9969 (1994) 24. T. Davenport, L. Zhou, J. Trivisono: Phys. Rev. B 59, 3421 (1999) 25. G.L. Zhao, B.N. Harmon: Phys. Rev. B 45, 2818 (1992); Phys. Rev. B 48, 2031 (1993) 26. S.M. Shapiro, B.X. Yang, Y. Noda, L.E. Tanner, D. Schryvers: Phys. Rev. B 44, 9301 (1991) 27. L. Ma˜ nosa, J. Zarestky, T. Lograsso, D.W. Delaney, C. Stassis: Phys. Rev. B 48, 15708 (1993) 28. P.-A. Ling˚ ard, O.G. Mouritsen: Phys. Rev. Lett. 57, 2458 (1986); Y.-Y. Ye, Y. Chen, K.-M. Ho, B.N. Harmon, P.-A. Lindg˚ ard: Phys. Rev. Lett. 58, 1769 (1987)

3 Disorder in Magnetic and Structural Transitions

47

29. B.E. Warren: X-Ray Diffraction (Dover Publications, New York 1990) p. 174 30. I.M. Robertson, C.M. Wayman: Phil. Mag. 48, 421 (1983); Phil. Mag. 48, 443 (1983); Phil. Mag. 48, 629 (1983) 31. S. Muto, S. Takeda, R. Oshima: J. Appl. Phys. 27, L1387 (1988) 32. L.E. Tanner, A.R. Pelton, R. Gronsky: J. Phys. 12, C4 (1982) 33. G. Van Tendeloo, M. Chandrasekaran, F.C. Lovey: Proceedings of the International Conference on Martensitic Transformations (The Japan Institute of Metals, Sendai, 1986), p. 868 34. V.A. Chernenko, C. Segu´ı, E. Cesari, J. Pons, V.V. Kokorin: Phys. Rev. B 57, 2659 (1998) 35. H. Seto, Y. Noda, Y. Yamada: J. Phys. Soc. 59, 978 (1990) 36. E. Cesari, V.A. Chernenko, V.V. Kokorin, J. Pons, C. Segu´ı: Acta Mater. 45, 999 (1997) 37. R. Kainuma, H. Nakano, K. Ishida: Metall. Mater. Trans. 27A, 4153 (1996) 38. This is a quite general characteristic in Cu-based shape memory alloys. See for instance [12] for a discussion 39. G. Fritsch, V.V. Kokorin, A. Kempf: J. Phys.: Condens. Matter 6, L107 (1994) 40. A. Zheludev, S.M. Shapiro, P. Wochner, M. Schwartz, M. Wall, L.E. Tanner: J. Phys. IV 5, C8 (1995) 41. K. Tsuchiya, A. Tsutsumi, H. Nakayama, S. Ishida, H. Ohtsuka, M. Umemoto: J. Phys. IV 112, 907 (2003) 42. S.M. Shapiro, B.X. Yang, G. Shirane, Y. Noda, L.E. Tanner: Phys. Rev. Lett. 62, 1298 (1989) 43. K. Ullakko, J.K. Huang, C. Kantner, R.C. O’Handley, V.V. Kokorin: Appl. Phys. Lett. 69, 1966 (1996) 44. L. Ma˜ nosa, A. Planes, J. Zarestky, T. Lograsso, D.L. Schlagel, C. Stassis: Phys. Rev. B 64, 24305 (2001) 45. U. Stuhr, P. Vorderwisch, V.V. Kokorin: J. Phys.: Condens. Matter 12, 7541 (2000) 46. U. Stuhr, P. Vorderwisch, V.V. Kokorin, P.A. Lindg˚ ard: Phys. Rev. B 56, 14360 (1997) 47. M. Stipcich, L. Ma˜ nosa, A. Planes, M. Morin, J. Zarestky, T. Lograsso, C. Stassis: Phys. Rev. B 70, 054115 (2004) 48. A. Gonz` alez-Comas, E. Obrad´ o, L. Ma˜ nosa, A. Planes, V.A. Chernenko, B.J. Hattink, A. Labarta: Phys. Rev. B 60, 7085 (1999) 49. A. Zheludev, S.M. Shapiro, P. Wochner, L.E. Tanner: Phys. Rev. B 54, 15045 (1996) 50. T. Cast´ an, E. Vives, P.A. Lindg˚ ard: Phys. Rev. B 60, 7071 (1999) 51. L. Ma˜ nosa, A. Gonz` alez-Comas, E. Obrad´ o, A. Planes, V.A. Chernenko, V.V. Kokorin, E. Cesari: Phys. Rev. B 55, 11068 (1997) 52. A. Planes, E. Obrad´ o, A. Gonz` alez-Comas, L. Ma˜ nosa: Phys. Rev. Lett. 79, 3926 (1997) 53. T. Cast´ an, A. Planes, A. Saxena: Phys. Rev. B 67, 134113 (2003) 54. V.K. Wadhawan. Introduction to Ferroic Materials (Gordon & Breach, UK, 2000) 55. R.D. Gomez. Experimental Methods in the Physical Sciences. Ed. by M. De Graef, Y. Zhu, Vol. 36 (2001), p. 69 56. M. De Graef. Experimental Methods in the Physical Sciences. Ed. by M. De Graef, Y. Zhu, Vol. 36 (2001), p. 27

48

T. Cast´ an et al.

57. Y. Murakami, D. Shindo, K. Oikawa, R. Kainuma, K. Ishida: Acta Mater. 50, 2173 (2002) 58. A. Saxena, T. Cast´ an, A. Planes, M. Porta, Y. Kishi, T.A. Lograsso, D. Viehland, M. Wuttig, M. De Graef: Phys. Rev. Lett. 92, 197203 (2004) 59. Z. Xu, M.C. Kim, J.F. Li, D. Viehland: Phil. Mag. A 74, 395 (1996) 60. S. Kartha, T. Cast´ an, J.A. Krumhansl, J.P. Sethna: Phys. Rev. Lett. 67, 3630 (1991) 61. H.S. Park, Y. Murakami, D. Shindo, V.A. Chernenko, T. Kanomata: Appl. Phys. Lett. 83, 3752 (2003) 62. M. De Graef, Y. Kishi, Y. Zhu, M. Wuttig: J. Phys. IV (Fr.) 112, 993 (2003) 63. Y. Kishi, M. De Graef, C. Craciunescu, T.A. Lograsso, D.A. Neumann, M. Wutting: J. Phys. IV (Fr.) 112, 1021 (2003) 64. O. Tikhomirov, H. Jiang, J. Levy: Phys. Rev. Lett. 89, 147601 (2002) 65. L. Yiping, A. Murthy, G.C. Hadjipanayis, H. Wan: Phys. Rev. B 54, 3033 (1996) 66. S. Sergeenkov, M. Ausloos: Phys. Rev. B 52, 3614 (1995) 67. M.M. Reyhani, P.G. McCornick: Scr. Metall. 21, 549 (1987) 68. J.P. Sethna, K.A. Dahmen, C.R. Myers: Nature 410, 242 (2001) 69. W. Cao, J.A. Krumhansl, R.J. Gooding: Phys. Rev. B 41, 11319 (1990); M. Rao, S. Sengupta: Phys. Rev. Lett. 78, 2168 (1997) 70. A. Planes, F.J. P´erez-Reche, E. Vives, L. Ma˜ nosa: Scr. Mater. 50, 181 (2004) 71. G. Bertotti: Hysteresis in Magnetism (Academic Press, San Diego, 1998) 72. V. Hardy, S. Majumdar, S.J. Crowe, M.R. Lees, D.M. Paul, L. Herv´e, A. Maignan, S. H´ebert, C. Martin, C. Yaicle, M. Hervieu, B. Raveau: Phys. Rev. B 69, 20407(R) (2004) 73. E. Dagotto, J. Burgy, A. Moreo: Solid State Commun. 126, 9 (2003) 74. K.H. Ahn, T. Lookman, A.R. Bishop: Nature 428, 401 (2004) 75. M.C. Mahato, S.R. Shenoy: J. Stat. Phys. 73, 123 (1993) 76. F.J. P´erez-Reche, M. Stipcich, E. Vives, L. Ma˜ nosa, A. Planes, M. Morin: Phys. Rev. B 69, 064101 (2004) 77. F. Casanova, A. Labarta, X. Batlle, E. Vives, J. Marcos, L. Ma˜ nosa, A. Planes, Eur. Phys. J. B 40, 427 (2004) 78. E. Puppin: J. Phys.: Condens. Matter 16, 1183 (2004) 79. G. Bertotti, G. Durin, A. Magni: J. Appl. Phys. 75, 5490 (1994) 80. F. P´erez-Reche, B. Tadi´c, L. Ma˜ nosa, A. Planes, E. Vives: Phys. Rev. Lett. 93, 195701 (2004) 81. K.A. Dahmen, J.P. Sethna: Phys. Rev. B 53, 14874 (1996) 82. R. Ahluwalia, G. Ananthakrishna: Phys. Rev. Lett. 86, 4076 (2001) 83. S. Sreekala, G. Ananthakrshna: Phys. Rev. Lett. 90, 135501 (2003)

4 Huge Magnetoresistance in Association with Strong Magnetoelastic Effects L. Morellon and M.R. Ibarra

There exists a broad phenomenology in solid state physics in which the interplay of interactions of different nature may be changed and controlled by external parameters such as temperature, magnetic field, and hydrostatic/chemical pressure. In some cases, this results in the modification of a delicate energetic balance in the system and, as a consequence, a phase transition or transformation can occur. In this review, we report significant magnetoresistance associated with strong magnetoelastic effects, as observed in intermetallic and magnetic oxide compounds.

4.1 Introduction Functional materials with a strong coupling between structural, magnetic, and electronic degrees of freedom are of prime interest nowadays because many different physical properties of both fundamental and applied interest, such as the magnetocaloric effect, magnetostriction, and magnetoresistance [1–5], are maximized. All the relevant phenomenology, which usually takes place in the vicinity of phase transitions, typically of first order, occurs in these systems upon changing external thermodynamic parameters, e.g., magnetic field, temperature, or hydrostatic pressure. For a general review on the recent advances and future directions in magnetic materials, the excellent review by Jiles [6] is recommended. In this contribution we focus on systems where an applied magnetic field brings about simultaneously a strong change in both the length of the sample (magnetostriction) and in the electrical resistivity (magnetoresistance). We present relevant examples of different materials exhibiting this distinct feature and the physical scenarios underlying the observed phenomenology will be described and discussed. Briefly, the most outstanding prototypes are systems undergoing one or more of the following phase transitions as a function of temperature: • • • •

Magnetic–crystallographic transformations Moment instabilities and spin fluctuations Metal–insulator transitions Charge/orbital instabilities.

50

L. Morellon and M.R. Ibarra

4.2 Magnetic–Crystallographic Transformations 4.2.1 Gd5 (Six Ge1−x )4 Gd5 (Six Ge1−x )4 is a unique family of intermetallic materials where many interesting properties and intriguing behavior have been recently discovered [7]. The unprecedented giant magnetocaloric effect [8–10], strong magnetoelastic effects [11, 12], and giant magnetoresistance [13–15] can be highlighted as the most relevant. This phenomenology has been associated with an intrinsically layered crystallographic structure combined with a magnetic–martensitic first-order phase transformation [16]. The coupled magnetic–crystallographic transition can be induced reversibly by the change of external parameters such as temperature, magnetic field, or applied pressure, and therefore, these alloys are attractive for their potential applications in magnetic refrigeration and/or as magnetic sensors and actuators [17]. Three extended solid solution regions exist in the temperature–composition (T –x) phase diagram [18,19]: the Si-rich solid solution, 0.575 ≤ x ≤ 1, has the orthorhombic Gd5 Si4 -type structure [O(I)]; the intermediate phase 0.4 < x ≤ 0.503 has a room temperature monoclinic (M ) structure; and the Ge-rich region, 0 < x ≤ 0.3 crystallizes in the Gd5 Ge4 -type orthorhombic structure [O(II)]. All three structures are composed of identical two-dimensional (2D) subnanometer-thick layers (slabs) interconnected via partially covalent interslab X–X bonds (X = Si, Ge). In the O(I) structure, all the slabs are interconnected by X–X bonds; half of these bonds are broken in the M structure and none remain in the O(II) structure. The magnetic–crystallographic transition involves breaking/reforming specific covalent X–X bonds [16] and the low-temperature ground state for all compositions 0 < x ≤ 1 is always ferromagnetic (FM) with all the slabs being interconnected, i.e., with the O(I) structure. The M structure is always paramagnetic (PM) whereas the O(II) can support either PM or antiferromagnetism (AFM) [12, 19]. The magnetic behavior of the R5 (Six Ge1−x )4 compounds can be understood qualitatively in terms of competition between intralayer (within the 2D slabs, conventional indirect 4f –4f RKKY) and interlayer exchange interactions (between slabs, direct Gd–Si/Ge–Gd superexchange propagated via the X–X bonds) [7, 20]. As an example we selected a Ge-rich alloy, Gd5 (Si0.1 Ge0.9 )4 . This compound shows upon cooling a second-order transition to the AFM state at TN = 127 K without changing the room-temperature O(II) structure, followed by a first-order magnetostructural transition O(II)[AFM] → O(I)[FM] at TC = 81 K. As seen in Fig. 4.1 the low-temperature O(I)[FM] state has a lower resistivity, ∆ρ/ρ ≈ 50%, and a larger length, ∆l/l ≈ 0.16% (along the particular direction represented in Fig. 4.1, a decrease as large as 1.6% being expected along the a direction [12]). Applying a magnetic field at T > TC induces a first-order transition into the O(I)[FM] state (Fig. 4.2) and, therefore, large magnetovolume and giant magnetoresistance effects [14] are expected as can be clearly seen in Fig. 4.3 at selected temperatures.

4 Huge Magnetoresistance in Association

51

Fig. 4.1. Zero-field electrical resistivity (ρ) and linear thermal expansion (∆l/l) of Gd5 (Si0.1 Ge0.9 )4 as a function of (increasing) temperature. The ρ values have been normalized to the 200 K value (ρ/ρ200 K ). The different crystallographic and magnetic phases are indicated

Fig. 4.2. Schematic representation of the crystallographic and magnetic structures of Gd5 (Si0.1 Ge0.9 )4 in the (a, b) plane at T < TN . Only the X = Si, Ge atoms participating in the X–X covalent-like bonds are depicted as solid spheres. A solid line linking the X atoms represents a formed bond [O(I)] whereas a dashed line is used for a broken one [O(II)]. Solid arrows are used to illustrate the change in the magnetic coupling as a function of magnetic field

4.2.2 MnAs Another well-known example of a system with a magnetostructural transition is MnAs. This alloy is ferromagnetic at room temperature and exhibits a firstorder loss of magnetic ordering during heating at TC ≈ 310 K, accompanied by a volume contraction of about 2% [21, 22]. This magnetic–crystallographic transformation involves a distortion from the FM high-symmetry hexagonal

52

L. Morellon and M.R. Ibarra

Fig. 4.3. Magnetoresistance ∆ρ/ρ (block symbols) and magnetostriction λ (open symbols) isotherms as a function of the applied magnetic field at selected temperatures of T = 100 K (squares) and 110 K (circles)

NiAs-type B81 phase (space group P 63 /mmc) into the lower symmetry PM orthorhombic MnP-type B31 phase (space group Pnma) [23]. At higher temperatures, TD ≈ 394 K, MnAs undergoes an additional second-order displacive transition back to the NiAs-type structure. The detailed magnetic phase diagram of MnAs was first established by Zieba et al. [24] and revised more recently by Chernenko et al. [25]. Magnetostriction values of 0.66% (linear contraction) were reported at the field-induced PM(B31) → FM(B81 ) transition in good agreement with the values determined from thermal expansion in zero field, see Fig. 4.4. Assuming isotropic behavior, the maximum volume magnetostriction is approximately equal to 2.1% at 310 K, nearly three times larger than the value associated to the AFM–FM transition in FeRh, see Sect. 4.3 and [26]. Recent neutron diffraction data in an applied field [27] give a direct microscopic evidence that the crystallographic transformation involves breaking and reforming metal–metal bonds. Coexistence and growth of the FM phase at the expense of the PM B31 phase as the magnetic field increases is observed, which is possible since the B31 phase is derived from the B81 phase by a cooperative displacement of pairs of [1, −1, 0] rows toward one another to form stronger Mn–Mn bonds. These also create shorter Mn–As bonds, which raise the antibonding states that σ-bond to the As atoms and trigger the transition to a low-spin state [28]. The field-induced phase transition is accompanied by an enhanced magnetoresistance response of about 17% at 310 K, see Fig. 4.5, in a similar fashion to that of the colossal magnetoresistance (CMR) Mn perovskites, as will be discussed later.

4 Huge Magnetoresistance in Association

53

Fig. 4.4. Linear thermal expansion as a function of temperature. The data points were measured in magnetic fields [25]

Fig. 4.5. Magnetoresistance ratio MR = {[ρ(0) − ρ(5 T)] /ρ(0)} × 100 versus temperature of MnAs. Inset: MR versus H [27]

4.3 Moment Instabilities and Spin Fluctuations 4.3.1 FeRh It is well known that in the ordered CsCl crystallographic structure of Fe0.5 Rh0.5 a PM–FM phase transition occurs at TC ≈ 650 K, within the FM phase a peculiar FM–AFM phase transition takes place at a temperature TF−AF ≈ 350 K [29]. Simultaneously, at this transition a 0.9% volume contraction occurs without change in the crystallographic structure together with a large drop in the electrical resistivity [30]. This singular transition

54

L. Morellon and M.R. Ibarra

strongly depends on the concentration, and it is only present in a very narrow concentration range of 5% around x = 0.5 in the binary Fe1−x Rhx phase diagram [31]. Neutron diffraction experiments [32] performed in the AF phase showed that only the iron atoms carry a relatively large local magnetic moment µFe ≈ 3.2 µB with no magnetic moment on the rhodium sites. In the FM phase, the iron moment remains with the same value, however, a local magnetic moment appears on the rhodium sites µRh ≈ 0.9 µB . First-principles total-energy band-structure calculations [33] predicted an equilibrium type-II AF magnetic ordering [antiferromagnetic coupling between successive layers of (111) iron layers] with no magnetic moment on the Rh sites. A metastable FM state is also predicted to exist with a larger unit cell volume than in the AF ground state. The FM state is very close in energy to the low-volume AF state and can be reached either thermally or by applying a magnetic field, as seen in magnetization measurements [26]. Thermal expansion measurements reveal that the stoichiometric FeRh alloy is an interesting system in which both Invar and anti-Invar effects are present in different range of temperatures [34]. The Invar-like behavior is observed at and below TC and is mainly attributed to transverse fluctuations of the local magnetic moments, which remain practically constant through and below the transition within the FM phase. The anti-Invar behavior is observed at TF−AF and is associated with the collapse of the local µRh , as a consequence of the longitudinal spin fluctuations at low temperatures. Huge field-induced volume magnetostriction (the anisotropic magnetostriction is negligible) has been observed in this alloy above room temperature, ω = (8.2 ± 0.2) × 10−3 [34]. As shown in Fig. 4.6, this value is reached above a certain critical field, above which it remains constant. The value observed for the field-induced volume magnetostriction is comparable to that found

Fig. 4.6. Isotherms of the volume magnetostriction ω for FeRh as function of the applied magnetic field (the arrows indicate the direction of the field variation) [34]

4 Huge Magnetoresistance in Association

55

Fig. 4.7. Thermal dependence of the critical field obtained from magnetostriction (filled circle) [34] and magnetoresistance (open circle) measurements [35]

in thermal expansion measurements and, therefore, the effect of the field is that of inducing a transition from the low-volume AF to the high-volume FM state through a first-order magnetic phase transition at a certain critical field. This gives experimental support to the theoretically predicted existence of a metastable ferromagnetic high-volume state in which Rh has a local magnetic moment. In addition, a large magnetoresistance ratio, ∆ρ/ρ ∼ = −50% has been found [35] at the field-induced transition in good agreement with that observed in zero field at the transition temperature. In Fig. 4.7 the temperature–magnetic field phase diagram as determined from both magnetoresistance and magnetostriction experiments is displayed, clearly demonstrating that both effects are coupled, and confirming earlier theoretical predictions [36] that proposed FeRh should behave as a natural multilayer system. 4.3.2 Hf1−x Tax Fe2 The Hf1−x Tax Fe2−y intermetallic compounds crystallize in the hexagonal MgZn2 -type structure (C-14 Laves phase, space group P 63 /mmc) with one Mg site (4f ) and two sites for Zn (2a and 6h). HfFe2 is a ferromagnet with an order temperature TC ≈ 600 K [37], and TaFe2 is a paramagnet [38]. The magnetic phase diagram of the Hf1−x Tax Fe2 series was reported by Nishihara and Yamaguchi [39–41], a first-order AFM–FM transition with decreasing temperature within the concentration range 0.1 < x < 0.3 being detected. In the FM state, the Fe ions in both sites carry a local magnetic moment of ≈1 µB , whereas in the AFM state only the ions at the 6h site have a local magnetic moment of ≈0.7 µB , with no magnetic moment being detected at the 2a sites. The fundamental question on whether the 2a Fe moment is identically zero, i.e., a nonlocal moment, or there is a local paramagnetic moment which vanishes on time averaging due to fast fluctuations within the experimental

56

L. Morellon and M.R. Ibarra

characteristic time has been answered in a recent contribution by means of M¨ ossbauer spectroscopy in an applied magnetic field [42]. The H–T phase diagram can be explained qualitatively in the framework of the Moriya and Usami theory for an itinerant electron system in which ferro- and antiferromagnetism coexist [43]. The thermally induced AFM–FM transition is accompanied by a large volume jump of ∆V /V ≈ 1.2% [40] and a resistivity anomaly of ∆ρ/ρ ≈ 25% (x = 0.2) [44]. The FM state can be reached from the AFM state by application of an external magnetic field and, therefore, strong magnetostriction and magnetoresistance effects are expected and indeed found experimentally [44–46]. In Fig. 4.8 the volume magnetostriction ω isotherms for Hf0.83 Ta0.17 Fe1.98 at selected temperatures are shown (the anisotropic magnetostriction is negligible). A huge value ω ≈ 0.7% close to room temperature is found. This value coincides with the zero-field thermal expansion, thus demonstrating this transition can be understood on the basis of a first-order transition from a low-volume low-spin state (AFM) to a high-volume high-spin state (FM), i.e., Invar behavior according to Moruzzi [47], and is attributed to the variation of the local moment of Fe at both sites of the crystallographic structure. The associated magnetoresistance effect in a Hf0.845 Ta0.155 Fe2 at various temperatures is displayed in Fig. 4.9. The observed giant magnetoresistance has been attributed to changes in the conduction electron scattering caused by the metamagnetic transition. 4.3.3 La(Fex Al1−x )13 The La(Fex Al1−x )13 alloys crystallize in the cubic NaZn13 -type structure in the composition range between x = 0.46 and x = 0.92. Such a dense structure

Fig. 4.8. Volume magnetostriction isotherms ω for Hf0.83 Ta0.17 Fe1.98 at some selected temperatures [46]

4 Huge Magnetoresistance in Association

57

Fig. 4.9. Magnetoresistance measured for Hf0.845 Ta0.155 Fe2 at various temperatures [45]

similar to fcc-Fe support different magnetic states due to the itinerancy of the 3d electrons, whose correlations bring about the complicated behavior of spin fluctuations [43]. The FM and AFM states are stabilized in the concentration range 0.60 ≤ x ≤ 0.86 and 0.87 ≤ x ≤ 0.92, respectively [48]. The FM and AFM states are very close in energy, and the AFM–FM metamagnetic transition takes place by application of magnetic field and pressure [48–50]. In addition, a temperature-induced first-order transition has been found in a narrow concentration range 0.860 < x < 0.869 [51]. As an example, the magnetoresistance and volume magnetostriction of the antiferromagnetic alloys x = 0.88 and x = 0.89, respectively, are depicted in Fig. 4.10. The magnetoresistance amounts to about 17% in a field of 4.76 T at low temperatures for x = 0.88. For x = 0.89, the volume magnetostriction at 4.2 K and above the spin-flip transition is huge, ω = 1%. The irreversibility at low fields seen in Fig. 4.10 (right) is due to the appearance of visible cracks in the sample. More information on the antiferromagnetic structure of these alloys can be found elsewhere [52].

4.4 Metal–Insulator Transitions The compounds that we describe are the mixed valent manganites, with formula R1−x Ax MnO3 where R3+ is a rare-earth ion and A2+ is an alkali earth. In the extreme compounds, RMnO3 and AMnO3 , the manganese ions are either Mn3+ (d4 ) or Mn4+ (d3 ). These compounds have crystallographic structures derived from the ideal cubic perovskite, the manganese ion occupying the center of oxygen octahedra. The crystal-electric-field splitting created by

58

L. Morellon and M.R. Ibarra

Fig. 4.10. Temperature dependence of the electrical resistivity ρ for an antiferromagnetic x = 0.88 alloy in zero field and in an applied magnetic field of 4.76 T greater than the critical field of the AFM–FM transition (left). Forced volume magnetostriction isotherm at 4.2 K for an AFM alloy x = 0.89 (right) [48]

the negative ligands imposes the symmetry of the magnetic electrons: t2g 3 eg 1 in the case of Mn3+ and t2g 3 in the case of Mn4+ . The magnetic interactions are governed by the superexchange antiferromagnetic interaction through the π-oxygen orbital t2g (Mn)–2pπ (O)–t2g (Mn). As a result, these compounds are antiferromagnets at low temperature. Substitution of R3+ by A2+ has a strong influence on the magnetic, structural, and magnetotransport properties of these compounds, leading to complex phase diagrams. The main effect is the creation of holes with eg symmetry. These electronic states are strongly hybridised with the σ oxygen orbitals in such a way that eg holes can easily hop from one Mn3+ ion to other Mn3+ with the only condition that the first Hund rule should not be violated (high spin state). Within this scheme the hole transfer is conditioned to the existence of eg occupied states with the same spin polarisation, i.e., the electronic transfer is favored under ferromagnetic alignment of the t2g spin (S = 3/2) cores. This corresponds to the double exchange interaction in which electron hopping over a ferromagnetic background gives rise to a very high conductivity. Then, ferromagnetism in these compounds is usually associated with a metallic behavior. The formalism to describe this interaction was proposed by Zener [53] and current models involve the use of ferromagnetic Kondo Hamiltonian to explain the relevant phenomena, see for instance the recent review article by Moreo et al. [54]. In general, the lack of long-range magnetic order or the existence of antiferromagnetic long-range order originated from the superexchange interaction are incompatible with the electron mobility, and the system is an insulator. Therefore, we can find very interesting systems in which the conduction electrons are fully polarized and show a high itinerancy under ferromagnetic

4 Huge Magnetoresistance in Association

59

alignment of the magnetic moments (“Half-metallic ferromagnets”). Because the hopping probability between neighboring Mn atoms is modulated by the alignment of the core t2g magnetic moments, large magnetoresistance can be expected. The optimum hole doping concentration is around x = 1/3, and for concentrations larger than x = 0.5, long-range ferromagnetic order is not present. In the paramagnetic phase a tendency toward localization takes place due to the electron–phonon interaction, which is quite relevant in these compounds as was demonstrated by the strong isotopic effect [55]. The interplay of the above mentioned interactions gives rise to a rich phenomenology reviewed in [56–58]. 4.4.1 Colossal Magnetoresistance (CMR) and Huge Volume Effects In this section we concentrate on the relevant magnetovolume effect discovered in the mixed-valent manganites. The measurements were performed on polycrystals by using the strain-gauge technique at low temperature and the push-rod method at temperatures higher than room temperature. These measurements compared quite well with the results obtained from neutron diffraction experiments. In Fig. 4.11 we show the thermal dependence of the resistivity and the thermal expansion for La0.6 Y0.7 Ca0.33 MnO3 . In the paramagnetic phase (T ≥ TC = 160 K) the resistivity has a semiconductor-like behavior. In this phase the compound is paramagnetic and insulator (PMI). Below TC the system is ferromagnetic and the resistivity sharply decreases by four orders of magnitude. This is a consequence of the double exchange interaction, which favors the electronic itinerancy. In this phase we found ferromagnetic and metallic resistivity (FMM). The observed dramatic effect on

Fig. 4.11. Thermal expansion and resistivity in La0.6 Y0.07 Ca0.33 MnO3 . An anomalous contribution over the phonon expansion above TC is observed. At this temperature, a sharp decrease in the resistivity takes place (metal–insulator transition)

60

L. Morellon and M.R. Ibarra

the resistivity also has a strong influence on the thermal expansion, which is manifested by a sharp decrease in the sample volume as one crosses from the PMI to the FMM phase [59] (the thermal expansion is isotropic). The Gr¨ uneisen fit of the thermal expansion curve using a Debye temperature θD = 500 K gives the results shown in Fig. 4.11, in which the thermal expansion in the FMM phase follows the phonon lattice behavior. However, in the PMI phase an extra anomalous contribution was found, being necessary to increase the temperature up to three times TC in order to recover the phonon behavior. Other interesting compound with the same hole concentration is the mixed valent manganite Pr0.7 Ca0.3 MnO3 [60]. The resistivity shows an insulating behavior even at low temperatures, which is a direct consequence of the absence of ferromagnetic order, and the thermal expansion results do not show any sharp anomaly, as observed in the former La–Ca compounds. Only a small bump at TCO = 210 K is observed, which is associated with the establishment of a charge ordering state. However, the thermal expansion does not follow the Gr¨ uneisen behavior. The extra volume contribution at low temperature is associated with the existence of an insulator antiferromagnetic state at low temperature below TN = 150 K. From these results one can easily conclude the existence of an extra volume contribution associated with the insulator state. The source of this volume anomaly was associated with a polaronic effect [59]. When an electron moves on an ionic background with an average lifetime shorter than 10−15 s (characteristic time associated with a 3d metal bandwidth) the wave function is characterized by a Bloch function. However, if the electron is near the ionic core for a period of ≈10−12 s, a polarization of the lattice is possible. The electric field created by the electron will attract the positive ions and push away the negative ions. The resultant electric field decreases because of the increase of the dielectric constant, due to the combined effect of polarization and ionic displacement. If the electron tries to move, it will feel a force that tends to restore it back to the original position. The result is the formation of a potential well originated by the electrostatic and elastic energy. The entity formed by the charge + distortion is called lattice polaron. Depending on the temperature range this polaron can move in the lattice, either by thermal activation overcoming the potential barrier or by tunneling forming a narrow polaronic band with a bandwidth given by the Heisenberg’s uncertainty relationship (∆ω∆τ =
). Recent NMR spin echo experiments [61] revealed that the average lifetime of these entities is ∆τp > 10−5 s. The extra contribution on the thermal expansion observed in the insulator regime is a consequence of this lattice polaron formation. The comparison of these results with small angle neutron scattering (SANS) gave evidence for the magnetic nature of the polaron responsible for the anomalous volume effect. One observes in a SANS experiment a magnetic contribution of ferromagnetic clusters with magnetic correlation length of the order 10–30 ˚ A [62]. The SANS intensity is related to the number of dispersion centers with short-

4 Huge Magnetoresistance in Association

61

range ferromagnetic correlations. By comparison with the thermal expansion anomaly, one can demonstrate how the volume anomaly associated with the localized charges or polarons has a magnetic origin. The magnetic correlation length diverges as TC is approached. This conclusion supported by SANS experiments can also be extracted from magnetostriction measurements. The magnetic polaron is the combination of a lattice polaron plus ferromagnetic clusters. This is a logical situation within the framework of the double exchange in mixed valent manganites. If the charge (electron or hole) is localized on a Mn ion, a probability of charge transfer to the next neighbor exists and this implies a parallel alignment or the Mn moments, i.e., the electronic hopping between surrounding Mn ions gives rise to the formation of a magnetic polaron with an average lifetime at a determined ionic site ∆τp ≥ 10−5 s, as determined from NMR experiments [61]. The same experiment also allowed to determine that the hopping time for the electron within the polaron is less than 10−9 which is the characteristic time given by the Larmor frequency for the indistinguishable Mn3+ –Mn4+ resonance line. 4.4.2 Magnetostriction and Magnetoresistance in the Paramagnetic Phase The magnetostriction measured in paramagnetic phase in mixed valent manganites with CMR is rather anomalous [59] with values, field and thermal dependence never observed before in a disordered magnetic phase. The Callen & Callen theory [63] using the one and two ion correlation functions predicts that the volume (ω) and anisotropic magnetostriction (λt ) in paramagnetic phase are proportional to the square of the applied magnetic field (ω = λ + 2λ⊥ ≈ H 2 and λt = λ − λ⊥ ≈ H 2 ) for localized magnetic moment systems. That is what is observed in many rare-earth intermetallics. The use of the crystal-electric-field theory allows using perturbation techniques to obtain information about the quadrupolar interaction in intermetallics [64]. This theory also predicts a quadratic dependence of parastriction with field. Usually the parastriction is very small and anisotropic in magnetic crystalline materials with well localized magnetic moments. We found in mixed valent manganites a very unusual behavior: the parastriction is huge, corresponds to a volume effect, and does not have a quadratic dependence with the field for temperaztures between TC < T < 1.8TC . In Fig. 4.12a we show the magnetostriction measured in the paramagnetic phase of La0.6 Y0.07 Ca0.33 MnO3 and, in Fig. 4.12b, we show the comparison with the anomalous contribution to the thermal expansion. The anisotropic magnetostriction was found to be negligible (λ ≈ λ⊥ ). In this compound the long-range magnetic order is destroyed at TC = 150 K as determined from the thermal dependence of the intensity of the neutron diffraction peaks. We can observe the large value of ω at fields larger than 2 T and a field dependence

62

L. Morellon and M.R. Ibarra

Fig. 4.12. Magnetostriction in the paramagnetic phase. (a) Volume magnetostriction isotherms. (b) Comparison of the extra contribution to the volume thermal expansion to the volume magnetostriction at 12 T. The lack of correspondence at temperatures well above TC is due to the lack of saturation of λt (H)

uncharacteristic of a paramagnet. This is, to our knowledge, the largest magnetostriction ever measured in a paramagnet, making this compound attractive not only from the point of view of the relevant magnetoresistance behavior, but also from the spectacular magnetostriction effect. Moreover, the tight relation between the magnetostriction and magnetoresistance offers a unique opportunity to get an insight into the relevant mechanisms operational in these compounds. The comparison of the anomalous spontaneous and high field value of magnetostriction shows that the field effect is the suppression of the volume anomaly in the thermal expansion. If we reconsider the previous argument, the observed magnetostriction in the paramagnetic phase should have its origin in the destruction of magnetic polarons under field, i.e., in the delocalization under field of the trapped charges. This argument would explain the sharp decrease of the resistivity under field (CMR). In addition

4 Huge Magnetoresistance in Association

63

to magnetostriction measurements, a microscopic study by means of SANS supported this assumption [62]. From the previous considerations we can describe the paramagnetic state of the CMR mixed valent manganite La–Ca as an insulating state in which localized charges give rise to an extra contribution to the spontaneous ω. These localized charges carry a ferromagnetic polarized cloud with a correlation length ξ ≈ 12 ˚ A. Under applied magnetic field ξ increases and the magnetic polarons percolate, inducing a ferromagnetic conducting state. Then, the huge magnetostriction observed in the paramagnetic phase originates from an insulator–metal transition, which is related to a percolative transition [62] from a high to a low volume state. We should call the attention of the readers to the peculiar and new behavior described here. First, the existence of well-defined nanometric ferromagnetic electronic cloud with an average life time larger than 10−5 s. This constitutes a kind of dynamic phase segregation known as magnetic polarons. They are responsible for the CMR and parastriction effects in manganites. Second, the magnetic polarons are of different nature than those proposed by Kasuya [65] and Nagaev [66] which are based on an impurity effect which does not give rise to magnetovolume effects. 4.4.3 Magnetostriction in Inhomogeneous Electronic Systems: Static Phase Segregation In Sect. 4.4.2 we underlined the relevance of dynamic phase segregation, in the form of magnetic polarons, on the magnetostriction in mixed valent manganites. In this section, we deal with a new phenomenology also found in these magnetic oxides. We show how magnetostriction experiments were essential to predict the existence of static phase separation, subsequently confirmed by other microscopic techniques. Nowadays, the existence of inhomogeneous states in different series of mixed valent manganites is well established [54]. The situation can be complex considering different doping level on the eg band as a result of x variation in compounds La1−x Mx MnO3 where M is an alkaline earth. In order to illustrate the magnetostriction effect when this phenomenon takes place, we selected a fixed hole concentration Mn4+ of 0.33, i.e., La2/3 Ca1/3 MnO3 . This compound is ferromagnetic below TC = 260 K and shows at this temperature a first-order transition [67] from the insulator state with formation of magnetic polarons to the metallic state, as we already described for La0.6 Y0.07 Ca1/3 MnO3 . The substitution of ≈10% of La by Tb has a strong effect on the ordering temperature, decreasing to TC = 150 K, but all the phenomenology is similar to that found in the sample with Y described previously. This is a general trend in manganites: the substitution of La3+ by another rare-earth ion as Tb3+ , keeping the same hole-doped concentration, produces a decrease of the ferromagnetic double exchange interaction strength. This is a result of the narrowing of the electronic eg bandwidth driven by the reduction of the Mn–O–Mn bond angle as a consequence of the smaller ionic radii of Tb3+ ions [68,69]. At x > 0.25 the compounds are not long-range magnetically

64

L. Morellon and M.R. Ibarra

Fig. 4.13. Volume thermal expansion for different compounds of the series La–Tb. Notice the low temperature behavior. The extra contribution to the thermal expansion persists for the insulating compounds. The metallic state follows the Gr¨ uneisen calculated phonon thermal expansion. The compound x = 0.25 remains in an intermediate state, which is an indication of the existence of an inhomogeneous state with static phase segregation (coexistence of FMM and PMI)

ordered and a spin glass state (disordered ferromagnetic clusters of ξ ≈ 20 ˚ A) is found in the region close to x = 0.25. At this particular concentration, long-range ferromagnetic ordered regions coexist with paramagnetic regions. First experimental evidence of phase segregation in this compound was obtained from the thermal expansion measurements. In Fig. 4.13 we show the thermal expansion results for x = 0.25 and other compounds of the series. At larger Tb doping level (x > 0.25) the systems are insulators at low temperatures and show similar behavior (i.e., high volume insulator state). However, for low Tb doping, x < 0.25, there exists a sharp anomaly associated with the volume reduction at TC , consequence of the insulator–metal transition as explained before. Unlike the two former systems, the considered compound x = 0.25 displays an intermediate volume state at low temperature. This state was considered as a result of an inhomogeneous para–ferromagnetic transition at TC ≈ 80 K [70]. The isothermal volume magnetostriction curves (not displayed here) do not saturate, which is interpreted as an indication of the stability of the inhomogeneous state even at 12 T. We can explain these results considering the establishment of microferromagnetic metallic regions below TC ≈ 80 K coexisting with a paramagnetic insulator even at very low temperature. Under field, the volume fraction of the ferromagnetic region increases with field, and consequently reducing the overall volume of the sample. These conclusions, directly obtained from magnetostriction measurements, were further confirmed by resistivity and SANS experiments. In Fig. 4.14 we show the temperature dependence of magnitudes of both quantities. We can observe a drop in the

4 Huge Magnetoresistance in Association

65

Fig. 4.14. Comparison of the resistivity with the SANS intensity at q = 0.1 ˚ A in x = 0.25. The field effect on the resistivity is also shown

resistivity at TC but the remnant low temperature resistivity is rather high (ρ ≈ 50 Ω cm). However, under applied field it decreases in two orders of magnitude. The final confirmation of the existence of an inhomogeneous state is the thermal dependence of SANS intensity. The enhancement of the SANS intensity as temperature decreases is the result of magnetic polaron formation as explained before. The sharp decrease of this magnitude at TC is an indication of formation of large clusters. These clusters are large enough to give a magnetic coherent contribution to the Bragg neutron diffraction intensity, as demonstrated in powder neutron diffraction experiments [70] (ξ > 500 ˚ A). This contribution is far from the magnetic signal expected from the establishment of long-range ferromagnetic order. Once more, the remanent SANS contribution at low temperatures can be attributed to the existence of shortrange ferromagnetic clusters.

4.5 Charge/Orbital Instabilities Charge ordering (CO) is a well-known phenomenon in magnetic oxides. The archetypical example is the magnetite Fe3 O4 . In this compound Verwey [71] proposed the long-range ordering of the Fe2+ and Fe3+ ions at the octahedral oxygen site of the spinel structure. This implies an extra charge periodicity which has been used as an example of Wigner crystal formation. This is a good example of strongly correlated states, i.e., the electrons in the real space try to keep away from each other as much as possible. The ground state is consequently insulating. Nevertheless, the exact nature and pattern of CO state in magnetite is a matter of discussion [72]. In mixed

66

L. Morellon and M.R. Ibarra

valent manganites the existence of CO has extensively been found in many compounds and is associated with the existence of orbital ordering. The static situation originates from the strong Coulomb repulsion. However, the ordered charges have eg symmetry with an anisotropic charge density on the ionic sites: 3z 2 − r2 , x2 − y 2 (Jahn–Teller degeneracy), this implies a new degree of freedom (pseudospin space) [73,74]. Of course, the space symmetry breaking of the pseudospin would make independent the charge and the orbital ordering. Nevertheless this situation is not observed in mixed valent manganites. The order of the charges alone would give rise only to an isotropic distortion of the unit cell. In spite of it, a strong anisotropic tetragonal distortion of the unit cell (without changing the symmetry) is found. A nice example is the compound La0.35 Ca0.65 MnO3 where superlattice electron diffraction spots indicated the long-range charge order [75] below TCO = 270 K. In addition, lattice distortions were detected by neutron diffraction experiments [76] suggesting the existence of an orbital ordering. The magnetic ordering is not always related to the existence of charge/orbital ordering. For instance, in this compound the long-range antiferromagnetic state occurs at TN ≈ 140 K, well below TCO [77]. Spin disorder or antiferromagnetism brings about charge localization. However, ferromagnetism causes delocalization through the double exchange mechanism. In this case the kinetic energy avoids the formation of the Wigner crystal. Transitions between charge ordering insulator and ferromagneticmetallic states can occur either spontaneously or under an applied field, when these two “quantum phases” have nearly degenerate minima in the free energy. We focus our attention on Pr0.7 Ca0.3 MnO3 . The crystallographic structure is orthorhombic, Pbnm space group. At TCO ≈ 210 K a transition to a charge ordering state takes place. Below this temperature the compound is an insulator down to 4 K. Nevertheless, a first-order insulator–metal transition can be induced by the application of a magnetic field. Below TN ≈ 150 K this compound orders antiferromagnetically in a CE-type structure. Below ≈100 K a canting of the magnetic moment was proposed giving rise to a ferromagnetic component [78]. A great effort was dedicated to establish the deep connection between magnetic and transport properties in this compound. Our effort was oriented toward the study of the lattice effects, with the aim of getting an insight into the underlying spin–charge–lattice coupling. In Fig. 4.15a we show the thermal expansion results on Pr0.7 Ca0.3 MnO3 under applied magnetic field of 0, 6 and 12 T. A comparison of the experimental results and the calculated lattice contribution (using a θD = 500 K) allows us to distinguish three different temperature ranges. Below ≈400 K there exists an extra contribution, as observed in the La–Ca compound, which we associated with the charge localization of small polarons. At TCO ≈ 210 K we observe a smooth anomaly which coincides with the temperature at which the charge ordering was observed in neutron diffraction experiments [78]. We proposed [60] at this temperature a transition from an incoherent small polaron regime above TCO to a coherent regime in which a charge ordered lattice is

4 Huge Magnetoresistance in Association

67

Fig. 4.15. Linear thermal expansion (a) and resistivity (b) at different field values of Pr0.7 Ca0.3 MnO3 . Inset in (a) is the extra anharmonicity associated with the insulator state

formed. We would like to underline the lack of volume effect (isotropic distortion) at this transition, yet the lattice is strongly distorted (large anisotropic change in the lattice parameters). Under an applied magnetic field of 6 T at ≈80 K a change in the volume is observed (≈0.14%). At 12 T the field induced transition is shifted up to 210 K. It apparently seems that at this field value, the charge ordering is completely suppressed in the whole temperature range. These results can be understood considering that the volume distortion associated with the charge localization (due to an extra anharmonic contribution) disappears as the charge is released. When that happens, the thermal expansion recovers the Gr¨ uneisen thermal dependence (phonon contribution). This is the situation of the metallic state in La2/3 Ca1/3 MnO3 . To make this effect more clear the extra anharmonicity is displayed in the inset of Fig. 4.15a.

68

L. Morellon and M.R. Ibarra

In this compound, as happens in the reported La–Ca compounds, a close correlation of the volume with the resistivity exists. As can be observed in Fig. 4.15b, the spontaneous behavior is characterized by a semiconductor-like dependence of the resistivity with a smooth anomaly at 210 K. Under a field of 6 T an insulator–metal transition takes place at ≈80 K and at 12 T the charge ordered state is not reached, in good agreement with thermal expansion measurements. The analysis of the magnetostriction and magnetoresistance allows an additional piece of information to be obtained. Two different behaviors were found above and below TCO as can be observed in Fig. 4.16a, b, respectively. The volume magnetostriction above TCO shows a continuous increase of volume with field and no hysteresis is observed. This behavior is similar to the already reported one for the La–Ca compound with the same carrier concen-

Fig. 4.16. Volume magnetostriction isotherms above (a) and below (b) the charge ordering transition

4 Huge Magnetoresistance in Association

69

tration. Thus, we associate this magnetostriction with the progressive delocalization of the carriers. The small value of this magnetostriction and the lack of saturation may be related to a weak magnetic character of the polaron. Fields higher than 14 T would be necesary in order to completely release the charge. The behaviour below TCO is completely different. Below 150 K the volume magnetostriction isotherms show a clear saturation with a very large hysteresis. This constitutes an indication of the complete destruction of the charge ordered state with the field, and also the first-order nature for the insulator–metal transition. The change in volume at the field-induced transition has the same value as the spontaneous extra volume contribution and opposite sign. A comparison of Fig. 4.17a, b shows the close relation between the magnetoelastic and magnetotransport properties in this compound. All the reported results on this compound reveal the different nature of the interplaying interactions in mixed valence manganites. Above TCO the Weiss temperature is positive [79] and the presence of lattice polarons with a weak magnetic character supports the presence of short-range ferromagnetic

Fig. 4.17. Comparison of the close correlation between magnetostriction and resistivity in Pr0.7 Ca0.3 MnO3

70

L. Morellon and M.R. Ibarra

correlations of double exchange origin. The existence of lattice polarons is an indication of the strong electron–phonon interaction. Below TCO the progressive localization of the carrier favors the ordering of the charges, probably by the long-range coulomb interaction. Below TN the compound orders antiferromagnetically due to the superexchange interaction. Then, it is not surprising to find a large field effect, given the different ground states available which are almost degenerate in energy. It is also worth mentioning some recent magnetovolume effects found in compounds in which the long-range charge/orbital ordering is not achieved, but the existence of short-range orbital order can lead to large magnetovolume effects when the system is close to an orbital instability, due to the above mentioned competition between ferrometallic and charge/orbital order insulator. This situation is realised in the series Sm1−x Srx MnO3 . This compound has para-insulator to ferromagnetic-metallic transition below TC = 120 K. At this transition a sharp decrease of the resistivity in more than three orders of magnitude was found and also large field induced strains [80]. To get an insight on the mechanisms underlying this behavior [81] we performed thermal expansion and magnetostriction measurements. The results can be explained in a similar way as in the former La2/3 Ca1/3 MnO3 compound described previously. The extra contribution to the thermal expansion was associated with the existence of magnetic polarons above TC . In these compounds, however, short-range orbital order has been observed from the analysis of the X-ray resonant scattering around the 220 Bragg reflection [74]. The extra contribution to the volume thermal expansion nicely scales with the intensity of the diffuse X-ray scattering, in a similar fashion to the scaling of the extra contribution to the thermal expansion above Tc in La2/3 Ca1/3 MnO3 with the SANS intensity. This short-range orbital ordering may be also associated with the formation of a ferromagnetic cloud and consequently with magnetic polarons. SANS experiments confirmed the coexistence of both kinds of short-range ordered regions [82]. With all the experimental evidence a polaronic behavior (Janh–Teller or orbital polarons) seems to be responsible for the localization of carriers which gives rise to the spontaneous magnetoelastic effects.

4.6 Conclusion In this short review, we selected a number of different archetypal compounds in which magnetoresistance and magnetostriction effects are intimately related. In general, the underlying mechanism is the electron itinerancy in magnetic oxides and, the change of the electronic structure due to a structural transformation or the modification of the magnetic moment due to a change of the longitudinal fluctuations. In spite of the vast phenomenology, a complete and unified theoretical description of the experiments is still lacking. The aim of this review has been to show that magnetostriction can provide relevant information on physical processes in solids and how the volume or anisotropic

4 Huge Magnetoresistance in Association

71

magnetostriction are at the base of magnetic or orbital instabilities which drastically influence the electrical transport properties. Acknowledgments Most of the reported work is based on experiments performed in our group. The authors would like to thank P.A. Algarabel, C. Marquina, J.M. De Teresa, C. Ritter, and Z. Arnold.

References 1. V.K. Pecharsky, K.A. Gschneidner Jr.: J. Magn. Magn. Mater. 200, 44 (1999) 2. A.M. Tishin: Magnetocaloric effect in the vicinity of phase transitions. In: Handbook of Magnetic Materials, ed. by K.H.J. Buschow. Vol. 12 (Elsevier, Amsterdam, 1999), pp. 395–524 3. A. Del Moral, P.A. Algarabel, J.I. Arnaudas, L. Benito, M. Ciria, C. de la Fuente, B. Garc´ıa-Landa, M.R. Ibarra, C. Marquina, L. Morellon, J.M. De Teresa: J. Magn. Magn. Mater. 242–245, 788 (2002) 4. A. Barth´el´emy, A. Fert, J.P. Contour, M. Bowen, V. Cros, J.M. De Teresa, A. Hamzic, J.C. Faini, J.M. George, J. Grollier, F. Montaigne, F. Pailloux, F. Petroff, C. Vouille: J. Magn. Magn. Mater. 242–245, 68 (2002) 5. V. Sechovsky, L. Havela, H. Nakotte, E. Bruck: IEEE Trans. Magn. 32, 4687 (1996) 6. D.C. Jiles: Acta Mater. 51, 5907 (2003) 7. V.K. Pecharsky, K.A. Gschneidner Jr.: Adv. Mater. 13, 683 (2001) 8. V.K. Pecharsky, K.A. Gschneidner Jr.: Phys. Rev. Lett. 78, 4494 (1997) 9. V.K. Pecharsky, K.A. Gschneidner Jr.: Appl. Phys. Lett. 70, 3299 (1997) 10. V.K. Pecharsky, K.A. Gschneidner Jr.: J. Magn. Magn. Mater. 167, L179 (1997) 11. L. Morellon, P.A. Algarabel, M.R. Ibarra, J. Blasco, B. Garc´ıa-Landa, Z. Arnold, F. Albertini: Phys. Rev. B 58, R14 721 (1998) 12. L. Morellon, J. Blasco, P.A. Algarabel, M.R. Ibarra: Phys. Rev. B 62, 1022 (2000) 13. L. Morellon, J. Stankiewicz, B. Garc´ıa-Landa, P.A. Algarabel, M.R. Ibarra: Appl. Phys. Lett. 73, 3462 (1998) 14. L. Morellon, P.A. Algarabel, C. Magen, M.R. Ibarra: J. Magn. Magn. Mater. 237, 119 (2001) 15. E.M. Levin, V.K. Pecharsky, K.A. Gschneidner Jr.: Phys. Rev. B 60, 7993 (1999) 16. W. Choe, V.K. Pecharsky, A.O. Pecharsky, K.A. Gschneidner Jr., V.G. Young Jr., G.J. Miller: Phys. Rev. Lett. 84, 4617 (2000) 17. D.C. Jiles, C.C.H. Lo: Sens. Actuators A 106, 3 (2003) 18. V.K. Pecharsky, K.A. Gschneidner Jr.: J. Alloys Compd. 260, 98 (1997) 19. A.O. Pecharsky, K.A. Gschneidner Jr., V.K. Pecharsky, C.E. Schindler: J. Alloys Compd. 338, 126 (2002) 20. E.M. Levin, V.K. Pecharsky, K.A. Gschneidner Jr.: Phys. Rev. B 62, R14 625 (2000) 21. B.T.M. Willis, H.P. Rooksby: Proc. Phys. Soc. Lond. Sect. B 67, 290 (1954)

72 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44.

45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57.

L. Morellon and M.R. Ibarra C.P. Bean, D.S. Rodbell: Phys. Rev. 126, 104 (1962) R.H. Wilson, J.S. Kasper: Acta Crystallogr. 17, 95 (1964) A. Zieba, Y. Shapira, S. Foner: Phys. Lett. 91A, 243 (1982) V.A. Chernenko, L. Wee, P.G. McCormick, R. Street: J. Appl. Phys. 85, 7833 (1999) J.B. McKinnon, D. Melville, E.W. Lee: J. Phys. C 1, S46 (1962) J. Mira, F. Rivadulla, J. Rivas, A. Fondado, T. Guidi, R. Caciuffo, F. Carsughi, P.G. Radaelli, J.B. Goodenough: Phys. Rev. Lett. 90, 097 203/1 (2003) J.B. Goodenough, J.A. Kafalas: Phys. Rev. B 157, 389 (1967) M. Fallot, R. Hocart: Rev. Sci. 77, 498 (1939) J.S. Kouvel, C.C. Hartelius: J. Appl. Phys. 33, 1343 (1962) G. Shirane, C.W. Chen, P.A. Flinn: Phys. Rev. 131, 183 (1963) G. Shirane, C.W. Chen, P.A. Flinn, R. Nathans: J. Appl. Phys. 34, 1044 (1963) V.L. Moruzzi, P.M. Marcus: Phys. Rev. B 46, 2864 (1992) M.R. Ibarra, P.A. Algarabel: Phys. Rev. B 50, 4196 (1994) P.A. Algarabel, M.R. Ibarra, C. Marquina, A. del Moral, J. Galibert, M. Iqbal, S. Askenazy: Appl. Phys. Lett. 66, 3062 (1995) V.L. Moruzzi, P.M. Marcus: Phys. Rev. B 46, 14 198 (1992) T. Nakamichi, K. Kai, Y. Aoki, K. Ikeda, M. Yamamoto: J. Phys. Soc. Jpn 29, 794 (1970) K. Kai, T. Nakamichi, M. Yamamoto: J. Phys. Soc. Jpn 29, 1094 (1970) Y. Nishihara, Y. Yamaguchi: J. Phys. Soc. Jpn 51, 1333 (1982) Y. Nishihara, Y. Yamaguchi: J. Phys. Soc. Jpn 52, 3630 (1983) Y. Nishihara, Y. Yamaguchi: J. Magn. Magn. Mater. 31–34, 77 (1983) H.R. Rechenberg, L. Morellon, P.A. Algarabel, M.R. Ibarra: Phys. Ref. B 71, 104412 (2005) T. Moriya, K. Usami: Solid State Commun. 23, 935 (1977) H.G.M. Duijn, E. Br¨ uck, A.A. Menovsky, K.H.J. Buschow, F.R. de Boer, R. Coehoorn, M. Winkelmann, K. Siemensmeyer: J. Appl. Phys. 81, 4218 (1997) Y. Nagata, T. Hagii, S. Yashiro, H. Samata, S. Abe: J. Magn. Magn. Mater. 292, 11(1999) L. Morellon, P.A. Algarabel, M.R. Ibarra, Z. Arnold, J. Kamarad: J. Appl. Phys. 80, 6911 (1996) V.L. Moruzzi: Phys. Rev. B 41, 6939 (1990) T.T.M. Palstra, G.J. Nieuwenhuys, J.A. Mydosh, K.H.J. Buschow: Phys. Rev. B 31, 4622 (1985) K. Irisawa, A. Fujita, H. Mitamura, T. Goto, K. Fukamichi: J. Alloys Compd. 327, 17 (2001) K. Irisawa, A. Fujita, H. Mitamura, T. Goto, K. Fukamichi: J. Alloys Compd. 329, 42 (2001) K. Irisawa, A. Fujita, K. Fukamichi: J. Alloys Compd. 305, 17 (2000) R.B. Helmholdt, T.T.M. Palstra, G.J. Nieuwenhuys, J.A. Mydosh, A.M. van der Kraan, K.H.J. Buschow: Phys. Rev. B 34, 169 (1986) C. Zener: Phys. Rev. 82, 403 (1951) A. Moreo, S. Yunoki, E. Dagotto: Science 283, 2034 (1999) G.M. Zhao, K. Conder, H. Keller, K.A. Muller: Nature 381, 676 (1996) C.N.R. Rao, B. Raveau: Colossal Magnetoresistance charge ordering and related properties of manganese oxides (World Scientific, Singapore, 1998) J.M.D. Coey, M. Viret, S. von Molnar: Adv. Phys. 48, 167 (1999)

4 Huge Magnetoresistance in Association

73

58. Y. Tokura: Colossal Magnetoresistive Oxides (Gordon and Breach, New York, 2000) 59. M.R. Ibarra, P.A. Algarabel, C. Marquina, J. Blasco, J. Garc´ıa: Phys. Rev. Lett. 75, 3541 (1995) 60. J.M. De Teresa, M.R. Ibarra, C. Marquina, P.A. Algarabel, S. Oseroff: Phys. Rev. B 54, R12 689 (1996) 61. C. Kapusta, P.C. Riedi, W. Kocemba, G.J. Tomka, M.R. Ibarra, J.M. De Teresa, M. Viret, J.M.D. Coey: J. Phys.: Condens. Matter 11, 4079 (1999) 62. J.M. De Teresa, M.R. Ibarra, P.A. Algarabel, C. Ritter, C. Marquina, J. Blasco, J. Garc´ıa, A. del Moral, Z. Arnold: Nature 386, 256 (1997) 63. E.R. Callen, H.B. Callen: Phys. Rev. 139, A455 (1965); E.R. Callen, H.B. Callen: Phys. Rev. 129, 578 (1963) 64. P. Morin, D. Schmitt: Quadrupolar interactions and magnetoelastic effects in rare-earth intermetallic compounds. In: Ferromagnetic Materials, ed. by K.H.J. Buschow, E.P. Wohlfarth. Vol. 5 (Elsevier, Amsterdam, 1990), pp. 1–132 65. T. Kasuya, A. Yanase: Rev. Mod. Phys. 40, 684 (1968) 66. E.L. Nagaev: Physi. Status Solidi b 65, 11 (1974); E.L. Nagaev: J. Magn. Magn. Mater. 110, 39 (1992) 67. J.M. De Teresa, M.R. Ibarra, J. Blasco, J. Garc´ıa, C. Marquina, P.A. Algarabel, Z. Arnold, K. Kamenev, C. Ritter, R. Von Helmolt: Phys. Rev. B 54, 1187 (1996) 68. J. Fontcuberta, B. Mart´ınez, A. Seffar, S. Pi˜ nol, J.L. Garc´ıa-Mu˜ noz, X. Obradors: Phys. Rev. Lett. 76, 1122 (1996) 69. J.M. De Teresa, M.R. Ibarra, C. Marquina, P.A. Algarabel, J. Blasco, J. Garc´ıa, A. Del Moral: Phys. Rev. Lett. 76, 3392 (1996) 70. J.M. De Teresa, C. Ritter, M.R. Ibarra, P.A. Algarabel, J.L. Garc´ıa-Mu˜ noz, J. Blasco, J. Garc´ıa, C. Marquina: Phys. Rev. B 56, 3317 (1997) 71. E.J.W. Verwey, P.W. Haayman: Physica 8, 979 (1941) 72. J. Garc´ıa, G. Sub´ıas, M.G. Proietti, H. Renevier, Y. Joly, J.L. Hodeau, J. Blasco, M.C. S´ anchez, J.F. B´erar: Phys. Rev. Lett. 85, 578 (2000) 73. K.I. Kugel, D.I. Khomskii: Usp. Fiz. Nauk 136, 621 (1982) [Sov. Phys. Uspekhi 25, 231 (1982)] 74. Y. Tokura, N. Nagaosa: Science 288, 462 (2000) 75. C.H. Chen, S.W. Cheong: Phys. Rev. Lett. 76, 4042 (1996) 76. M.R. Ibarra, J.M. De Teresa, J. Blasco, P.A. Algarabel, C. Marquina, J. Garc´ıa, J. Stankiewicz, C. Ritter: Phys. Rev. B 56, 8252 (1997) 77. H. Kuwahara, Y. Tokura: First-order insulator–metal transitions in perovskite manganites with charge-ordering instability. In: Colossal Magnetoresistance Charge Ordering and Related Properties of Manganese Oxides, ed. by C.N.R. Rao, B. Raveau (World Scientific, Singapore, 1998), pp. 217–239 78. H. Yoshizawa, H. Kawano, Y. Tomioka, Y. Tokura: Phys. Rev. B 52, 13145 (1995) 79. M.R. Lees, J. Barratt, G. Balakrishnan, D.M.K. Paul, M. Yethiraj: Phys. Rev. B 52, 14303 (1995) 80. Y. Tomioka, H. Kuwahara, A. Asamitsu, M. Kasai, Y. Tokura: Appl. Phys. Lett. 70, 3609 (1997) 81. C. Marquina, M.R. Ibarra, A.I. Abramovich, A.V. Michurin, L.I. Koroleva: J. Magn. Magn. Mater. 226, 999 (2000) 82. J.M. De Teresa, M.R. Ibarra, P. Algarabel, B. Garc´ıa-Landa, C. Marquina, C. Ritter, A. Maignan, C. Martin, B. Raveau, A. Kurbakov, V. Trounov: Phys. Rev. B 65, 100403 (2002)

5 Interplay of Spin, Charge, and Lattice in CMR Manganites and HTSC Cuprates T. Egami

5.1 Introduction In the field of electronics technology oxides have been primarily used only as insulators, such as native SiO2 , or as magnetic materials, for instance ferric oxides as recording media and recording heads. For semiconductors doped silicon has been used almost exclusively, and simple metals, such as aluminum, copper and gold, have been used for contacts and connectors. This practice, however, could change dramatically in the next decade or so. Firstly, hightemperature superconductivity (HTSC) was discovered in the cuprates by Bednorz and M¨ uller [1]. This was a major surprise, since majority of oxides are insulators or poor conductors. Then the so-called colossal magnetoresistivity (CMR) was observed in some manganites [2]. These discoveries could be harbingers for oxides to become mainstream electronic materials in the future. These functional oxides are not simple oxides such as SiO2 or Al2 O3 , but are complex oxides composed of many elements including transition metals, and usually have a complex lattice structure. Not only is the lattice structure complex but also the physical interactions in these solids are complex, involving spin, charge, orbital, and lattice degrees of freedom. The remarkable properties of these complex oxides originate from the interplay, i.e., competition and collaboration, among these multiple degrees of freedom. Such a state of strongly interacting forces was rarely encountered in the history of the condensed matter research, where most phenomena were explained in terms of the linear response to small perturbation. Understanding of these competing forces requires the use of a multitude of novel experimental techniques and new theoretical approaches, usually involving extensive numerical simulations. A major difficulty often faced in studying these intertwined multiple degrees of freedom is that complexity of the phenomena often conceals the true order parameter. In particular, disorder in the system tends to have much more profound effects than in simple systems, leading to localization of the effects and destroying the long-range order. Thus the conventional experimental

76

T. Egami

and theoretical approaches that rely on the long-range order are frequently ineffective, and we need local approaches to separate intertwined forces. In this chapter we describe how the pulsed neutron atomic pair-density function (PDF) approach has facilitated understanding of the mechanism of the ionic size effect on the CMR manganites and the spin–charge stripes in the cuprates, and how PDF measurements and inelastic pulsed neutron scattering measurements of phonons revealed involvement of phonons in the HTSC phenomenon.

5.2 Stability of Polaronic Phase in the CMR Manganites 5.2.1 Colossal Magnetoresistivity Certain transition metal oxides such as manganites A1−x A
x MnO3 , where A is a trivalent ion (rare earths) and A
is a divalent ion (alkali earth or Bi and Pb), undergo a phase transition from a paramagnetic or antiferromagnetic insulator to a ferromagnetic metal. Near the point of the metal–insulator transition (MIT), an applied magnetic field can induce the ferromagnetic transition and thus can greatly change the resistivity. This effect, called CMR [2], was extensively studied during the last ten years as a representative case of unusual properties of complex oxides fueled by the possibility of application. The basic properties of the oxides involved have been known for a long time, going back to the pioneering study by Jonker and Van Santen [3]. Initially the MIT in the manganites was considered to be caused purely by the magnetic interaction; metallic conduction is possible only in the ferromagnetic state, while in the paramagnetic state spin scattering causes high resistivity and creates spin polarons. This simple view was first questioned by Millis et al. [4] who advocated the need for considering other interactions to confine and localize charge carriers as polarons, most likely the lattice contribution involving the Jahn–Teller (JT) effect, to fully explain the experimental results. The lattice contribution to polarons was soon experimentally confirmed by various local probes [5–8]. For instance a Mn3+ ion in insulating LaMnO3 has four d-electrons in the t2g 3 -eg 1 high-spin configuration. Because of the singly occupied eg level this ion is JT active, and the MnO6 octahedron becomes elongated along one axis. In this case there are two Mn–O bond lengths, four short (∼1.95 ˚ A) and two long (∼2.1 ˚ A) within the MnO6 octahedron. On the other hand a Mn4+ ion has no eg electron, so that the MnO6 octahedron retains cubic symmetry and all the six Mn–O neighbor distances are the same. If doped holes in La1−x Cax MnO3 are localized to form polarons they affect the lattice structure by locally reducing the JT distortion. But unless polarons form a long-range order such changes in the local JT distortion cannot be easily detected by the standard diffraction methods. On the other hand the local structural probes, such as the atomic PDF analysis of neutron diffraction or the extended X-ray absorption fine structure (EXAFS) method, convincingly

5 Interplay of Spin, Charge, and Lattice in Manganites and Cuprates

77

demonstrated that the local JT distortion exists at some Mn sites but not at others. The Mn sites without or reduced JT distortion must coincide with the sites with localized holes.

5.2.2 Pulsed Neutron PDF Method The structure of a crystal is defined by its symmetry, lattice parameters, and atomic positions within the unit cell, all of which can be determined by measuring the position and the intensity of the Bragg diffraction peaks. Thus usually structure determination implies the analysis of the Bragg peaks in the diffraction pattern. However, real crystals are never perfect and contain lattice defects. In complex systems such as the CMR oxides electron–lattice interaction causes local changes in the lattice structure and destroys translational symmetry, but such local changes are not directly reflected in the crystal structure. These short-range deviations from the perfect lattice structure produce diffuse scattering [9]. If the diffuse scattering forms a broad peak, the position of the peak indicates the periodicity of the superstructure and the width gives the information on the structural correlation length. But in order to determine the actual deviations, for instance in terms of atomic displacements, many diffuse scattering peaks have to be studied and analyzed. While such studies are possible and useful information has been obtained [10], a simpler and more reliable approach is to use the PDF analysis. In the PDF analysis the powder diffraction intensities are determined with neutrons and X-rays with relatively high energies up to a large momentum transfer, Q (=4π sin θ/λ, θ is the diffraction angle and λ is the wavelength of the scattering probe), and the results are Fourier-transformed to obtain the atom–atom correlation function after correction for absorption and other factors [11]. The pair-density function, ρ(r), is defined by ρ(r) = ρ0 g(r) =

 1 δ(r − rij ) , 2 4πN r i,j

(5.1)

where rij is the distance between atoms i and j, ρ0 is the number density of atoms, and N is the number of atoms in the system. It is customary to call g(r) the pair-distribution function, and 4πr2 ρ0 g(r) the radial-distribution function (RDF). They are obtained from the diffraction data by 1 ρ(r) = ρ0 + 2 2π r



∞

[S(Q) − 1] sin(Qr) Q dQ ,

(5.2)

0

where S(Q) is the normalized structure function determined by powder diffraction measurement. In powder diffraction measurement all the diffuse scattering intensities are measured in addition to the Bragg peaks, while in the case of direct

78

T. Egami

Fig. 5.1. The Neutron Powder Diffractometer (NPDF) of the Lujan Center, Los Alamos National Laboratory, during the upgrading work supported by the National Science Foundation DMR00-76488. A large backscattering detector frame for position-sensitive 3 He detectors is seen with a technician inside

measurement of diffuse scattering from a single crystal usually only a few peaks are chosen for measurement, since the diffuse scattering intensities are usually weak and the measurement is time consuming. Thus the PDF represents full atomic correlation, not compromised by shortcuts in the experiment. This is an important point, since it is so easy to miss critical information by not studying all the diffuse intensities. Even though the angular information is lost by powder averaging the PDF method is a valuable tool that should be used first in evaluating the local disorder. The PDF method has long been used almost exclusively for the study of short-range order in liquids and glasses, but the advent of modern radiation sources, such as pulsed neutron sources and synchrotron radiation sources, which deliver short wavelength probes, made it possible to apply this technique on crystalline solids with strong disorder [11]. Also by the use of high Q-resolution spectrometer, such as the Neutron Powder Diffractometer (NPDF) of the Lujan Center, Los Alamos National Laboratory (Fig. 5.1), it is now possible to determine the PDF up to 200 or 300 ˚ A as shown in Fig. 5.2, thereby seamlessly connecting the short-range structural analysis with the conventional analysis of the long-range order [12, 13]. This method has been quite effective in characterizing the local structure of CMR manganites and related oxides, and facilitating understanding of their physical properties in terms of their local structure [11, 14].

5 Interplay of Spin, Charge, and Lattice in Manganites and Cuprates

79

Fig. 5.2. The pair-density function (PDF) of crystalline Ni powder at T = 10 K obtained by the NPDF [13]

5.2.3 Polaron Stability While the polaron formation is central to the phenomenon of MIT in the manganites, it soon became recognized that the highly collective nature of the ground state is necessary to explain the high response to the magnetic field by the CMR manganites. In the insulating phase charges are not only confined in the polaronic state, but also form a charge-ordered state. CMR is achieved in the mesoscopic mixed phase state of charge-ordered insulator and metal [15]. Note that charge ordering and polaron formation are actually very closely related phenomena, since the charge-ordered state is simply the crystallized state of polarons (polaron lattice), rather than the charge-density waves (CDW) state driven by the Fermi surface instability. The competing forces that determine the stability of polarons, thus MIT, are the following: 1. Forces to localize charges; electron–phonon (e–p) interaction and spin correlation 2. Forces to delocalize charges; electron kinetic energy and lattice elasticity In classical mechanics a localized static body has a lower kinetic energy, but in quantum mechanics localization results in strong local variation in the wavefunction, and thus in a higher kinetic energy. Thus electron kinetic energy prefers charge delocalization. The balance between the two forces could be expressed by the renormalized e–p coupling constant, λ=

g2 , Ktcos θ2

(5.3)

where K is the relevant elastic constant, g is the bare e–p coupling parameter, t is the hopping integral in the tight-binding approximation and θ is the angle made by neighboring spins [16, 17]. If λ exceeds a critical value charges will be localized as polarons in the charge-ordered state. The value of λ varies with composition, and even when the charge density is the same, it depends on the ionic size of the A-site ions, rA [18]. When the

80

T. Egami

value of rA is large, the system is a stable ferromagnetic metal for x < 0.5, and when it is small it is an insulator. At the crossover state the solid is precariously balanced and the CMR behavior is observed. Therefore understanding this ionic size effect is crucial in understanding the forces that cause the CMR effect. It is generally considered that the balance is affected through the change in t, thus the electronic bandwidth [18, 19]. A large A-site ion expands the lattice and makes the Mn–O–Mn bond straight, while a small A-site ion contracts the lattice and makes the Mn–O–Mn bond buckled. In the buckled Mn–O–Mn bond the p–d overlap is reduced and the value of t is reduced accordingly. While this explanation sounds reasonable, a closer look reveals a fault in this explanation. Actually the change in the buckling angle of the Mn–O–Mn bond due to the ionic size change is rather small, and the expected change in t is too small to explain the observed drastic effect [20,21]. It is true that the manganites with small A-site ions have a narrow band, but that is a consequence of charge localization, and not the cause, because polaron formation renormalizes the electron bandwidth as shown by Holstein [22]. The true story is a bit more complex and involves the elastic field around a polaron [23, 24]. A lattice polaron is formed by displacements of neighboring ions that trap the charge carrier. In the case of the manganites these displacements are related to the JT distortion. The polaron site should have no JT distortion and has six short Mn–O bonds, while the insulating sites outside are JT distorted and have four short bonds and two long bonds as for Mn3+ . Thus the fraction of the sites without the JT distortion can be deduced by comparing the density of short Mn–O bonds and that of long Mn–O bonds. They can be determined by the PDF analysis. As shown in Fig. 5.3 Mn–O bonds are seen as split peaks in the neutron PDF. The peaks are negative since the neutron scattering length of Mn is negative. By evaluating the areas of these peaks the number ratio of long and short Mn–O bonds can be obtained, and the fraction of the sites with or without the local JT distortion can be calculated from the ratio. Figure 5.4 shows the fraction of the sites without the JT distortion as a function of the A-site ionic size, rA , evaluated by the pulsed neutron PDF study for A0.7 A
0.3 MnO3 [24]. By dividing the density of sites without the local JT distortion by the charge density (x = 0.3) it is possible to evaluate the size of the polaron. Figure 5.4 shows that when the ionic size is small polarons occupy a single site, and as the ionic size is increased the polaron expands, and finally practically all the sites lose local JT distortion and the system becomes a homogeneous metal. If the value of rA is large and the Mn–O–Mn bond is straight this local lattice contraction associated with the loss of local JT distortion produces a long-range strain field around it as shown in Fig. 5.5a. But if the Mn–O–Mn bond is sufficiently buckled when rA is small, the local lattice contraction is accommodated by the unbuckling of the bond as in Fig. 5.5b. In other words transverse phonons can screen the local strains. The effect of the longrange strain field on the polaron formation energy was evaluated using the

5 Interplay of Spin, Charge, and Lattice in Manganites and Cuprates

81

˚, Fig. 5.3. The RDF of Y0.5 La0.2 Ca0.3 MnO3 in the range of r from 1.7 to 2.3 A showing the peak for the Mn–O bonds, determined by pulsed neutron diffraction [23]. Here the RDF = 4πr2 ρ(r), where ρ(r) is the PDF and b is the composition average of the neutron scattering length. The Mn–O peak can be fit by two sub-peaks corresponding to short and long Mn–O bonds

Fig. 5.4. The fraction of the Mn sites without the local Jahn–Teller distortion in A0.7 A0.3 MnO3 with various A and A ions, as a function of the average ionic size, rA , determined by the PDF analysis of the pulsed neutron diffraction data [24]. It gives the size of the polaron when divided by the charge density in terms of the number of Mn sites involved

82

T. Egami

Fig. 5.5. The mechanism of local accommodation of polaronic strain. When the Mn–O–Mn bond is straight, loss of local Jahn–Teller distortion reduces the long (L) Mn–O bond to a short (S) one and creates a long-range stress field (a). However, when the Mn–O–Mn bond is sufficiently buckled, unbuckling locally accommodates the polaronic strain, and the long-range stress field will not be created (b) [23]

Fig. 5.6. The effective electron–lattice coupling parameter normalized to the singlesite value without the long-range stress field. The effect of the long-range stress field is to reduce the effective coupling parameter, since the energy to create a polaron is increased. The dashed line is for a slightly different choice of critical parameters [23]

continuum approximation by Eshelby [23]. It was found that the presence or absence of the long-range strain field changes the effective coupling constant λ by nearly a factor of 2, as shown in Fig. 5.6 [23]. In other words what changes with the A-site ionic size is not the value of t, but the effective value of K. This example shows that the details of the local atomic structure can have profound effects on the properties of complex oxides by affecting the balance among the competing forces.

5 Interplay of Spin, Charge, and Lattice in Manganites and Cuprates

83

Fig. 5.7. Dependence of the superconductive transition temperature, TC , on the ionic size of the site for La/Nd and Sr/Ba due to the formation of the stripe state [23, 26, 27]

5.3 Stability of Spin–Charge Stripes in the Cuprates In some cuprate compounds doped charges are localized in the spin–charge stripes composed of charged stripes and antiferromagnetic stripes [25]. The stability of the stripes also depends on the ionic size as well as charge density [25], and the size effect can be explained by the same mechanism just discussed for the CMR manganites [23]. For example, in the system (La,Nd)1.88 (Sr,Ba)0.12 CuO4 with the hole density x 1/8 the superconducting critical temperature, TC , depends strongly on the ionic size as shown in Fig. 5.7, and the stripes are observed when TC is suppressed, suggesting that the stripes and superconductivity compete against each other [26, 27]. Doped holes enter the hybridized Cu(dx2 −y2 )–O(px ) anti-bonding σ-orbital, thus when holes are localized they reduce the local Cu–O bond distance. Indeed the Cu–O distance in the averaged lattice structure changes from A in La1.85 Sr0.15 CuO4 [29] and 1.882 ˚ A in 1.903 ˚ A in La2 CuO4 [28] to 1.889 ˚ A for each La1.8 Sr0.2 CuO4 [30], all at 10 K, indicating contraction of about 0.1 ˚ hole. Since the linear hole density in the stripe is 50%, if all the holes are in the single row of oxygen or copper the Cu–O bond will contract by 0.05 ˚ A. A, and Figure 5.7 shows that the optimum state is achieved for rA = 1.237 ˚ A the stripe structure becomes stable. This when rA is reduced by 0.07 ˚ change in the La/Nd–O distance corresponds to the change in the Cu–O distance of 0.05 ˚ A, exactly the amount expected. Thus if this much of a change in the Cu–O distance can be accommodated by the unbuckling of the Cu–O– Cu bond, the stripe state can be stabilized. If the Cu–O–Cu bond is nearly straight and the contraction of the Cu–O bond creates a long-range stress field, the cost of the elastic energy to create the stripe state suppresses the stripes A [31]) and superconductivity remains alive. When a large Ba ion (rA = 1.47 ˚

84

T. Egami

is added instead of Sr, it expands the Cu–O–Cu bond locally, allowing it to become bifurcated and stabilizing the stripes. This analysis demonstrates that the same strain accommodation mechanism explains both the CMR phenomenon in the manganites and the stability of the stripes in the cuprates.

5.4 Electron–Phonon Coupling and Mechanism of Superconductivity in the Cuprates 5.4.1 Local Lattice Distortion and Inhomogeneous Electronic States For a long time the majority opinion on the microscopic mechanism of HTSC in the cuprates has been that it is a purely electronic phenomenon, probably described by the t–J Hamiltonian, and the lattice has no role [32]. However, recent evidences [33–35] are seriously beginning to challenge this long-held opinion by the majority. Indeed a large number of studies indicate that the lattice of the cuprates has local distortion, which changes near the superconducting transition temperature, TC , or the pseudo-gap temperature, TPG [36,37]. It is even possible that the lattice plays a crucial role in the HTSC phenomenon. Local structural probes suggest that below a certain temperature the local environment of Copper ion appears to show two distinct configurations, possibly corresponding to fully doped and strongly underdoped states [38,39]. The appearance of these two components is consistent with the presence of spin– charge stripe structure, but the evidence for the existence of such a structure is weak. The earlier discussion on the stripes suggests that unless the ionic size is favorable strong charge localization into the stripe structure is not likely to take place. It is more likely that weaker localization into some nanoscale domains with different charge densities is taking place. Indeed beautiful scanning tunneling microscopy–scanning tunneling spectroscopy (STM–STS) observations [34,35] revealed the presence of such nano-scale electronic inhomogeneity. While there are still some controversies regarding the STM results on the effect of surface, it is now generally accepted that in the cuprates the presence of electronic inhomogeneity does not interfere with superconductivity for some mysterious reason. However, not everybody agrees that the electronic inhomogeneity is necessary for superconductivity to occur. An experiment that would give an answer to this point will have an enormous impact. 5.4.2 Electron–Phonon Coupling in the Cuprates The observation of local lattice distortion associated with electronic inhomogeneity obviously implies coupling of the lattice with the electrons, but it could merely be a consequence of electronic inhomogeneity, rather than the cause. On the other hand more direct evidence of strong e–p coupling was found in the angle-resolved photoemission spectroscopy (ARPES) study [33].

5 Interplay of Spin, Charge, and Lattice in Manganites and Cuprates

85

It is particularly important that the ARPES result showed that the e–p coupling is strong only for a particular branch of the phonons, the Cu–O bond-stretching longitudinal optical (LO) modes toward the zone-boundary half-breathing mode (Fig. 5.8) that shows strong softening with doping and anomalous temperature dependence [40–44]. Inelastic neutron scattering study of the phonons in YBa2 Cu3 O6.95 has shown that a shift in the spectral intensity of scattering occurs with temperature [44]. In the energy range related to the Cu–O bond-stretching LO modes intensities are transferred from the range around 62 meV to the range around 53 meV as temperature is reduced. To illustrate this change the inelastic neutron scattering intensity was measured at various temperatures at Q = (3.25, 0, 0) in the units of reciprocal lattice vectors, or q = (π/2, 0, 0) in the phonon vector in the units of inverse lattice parameter. Figure 5.9 shows the difference in the average intensity from 51 to 55 meV, I(1), and from 56 to 68 meV, I(2), as a function of temperature [44]. The change occurs below about 90 K, close to TC = TPG (=93 K). For YBa2 Cu3 O6.95 (TC = 60 K) a similar change starts around 150 K ( TPG ), so that the change must be related to the pseudo-gap temperature, rather than TC .

Fig. 5.8. In plane Cu–O half-breathing mode at q = (π, 0)

Fig. 5.9. Difference in the average inelastic neutron scattering intensity from a single crystal of YBa2 Cu3 O6.95 at Q = (3.25, 0, 0), from 51 to 55 meV, I(2), and from 56 to 68 meV, I(1), as a function of temperature [44]. TC is 93 K for this sample

86

T. Egami

Fig. 5.10. Optical conductivity of the Cu–O chain with and without doping (x = 0, 0.25) and with and without e–p coupling (α = 0, 1) for the phonon with q = 0.25 and 0.5, calculated for a one-dimensional Hubbard model [48]. Note that the effect of phonon is found up to a large energy range

Such a strong e–p coupling of the Cu–O bond-stretching mode was theoretically predicted [45,46]. In simple metals the lattice vibration can be described in terms of the deformation potential, which is almost totally screened by free electrons, except for the Friedel oscillation. In the cuprates the screening length is longer than the interatomic distances because of low charge density, and Cu and O maintain their ionicity. Thus the d-orbitals of Cu and the p-orbitals of O move with the nuclei, so that the change in the interatomic distance between Cu and O due to lattice vibrations changes the p–d overlap. This changes the hopping integral, t, between Cu and O, in the mechanism known as the Su–Schrieffer–Heeger (SSH) coupling [47]. For simple metals the effect of phonons is limited to the electrons close to the Fermi level within the phonon frequency ωph . However, in the cuprates the effect extends up to many eVs away from the Fermi level, since the whole band structure is modified by the phonons. Consequently the optical conductivity is modified up to high energy scale as shown in Fig. 5.10 as calculated with the Hubbard model with the SSH coupling [48]. It was found that the optical conductivity is modified up to a large energy scale by superconductive transition and this was argued to be the evidence for an electronic mechanism [49–51]. However, this result suggests that the change in the optical conductivity with the onset of superconductivity is perfectly consistent with the phonon mechanism. It is interesting to note that in ferroelectric oxides, such as BaTiO3 , this SSH coupling produces electronic polarization as large as the ionic polarization [52, 53]. In the ionic state Ti4+ and O2− have d0 and p6 orbital occupation, but upon forming a covalent bond they hybridize to form a p–d orbital. Ti–O

5 Interplay of Spin, Charge, and Lattice in Manganites and Cuprates

87

Fig. 5.11. The q dependence of the effective charge of oxygen in the 1-d Hubbard model due to the LO phonon mode, calculated for the ring of N = 12 sites with doping level x = 0, 1/3 and for N = 16 with x = 0, 1/4. The dashed line indicates the ionic value (static charge) [48]

bond-stretching phonons modify this p–d overlap, thus t between Ti and O. If the Ti–O distance is reduced the p–d overlap will increase, and electron will be transferred from O to Ti. The current due to the charge transfer produces the electronic polarization [52, 53]. In the doped cuprates, on the other hand, reducing the distance between Cu and O results in the transfer of holes, rather than electrons, and the resulting electronic polarization has an opposite sense as the ionic polarization as shown in Fig. 5.11 [48]. It is noted that for the titanates the electronic polarization is maximum at q = 0, resulting in uniform ferroelectric polarization, but in the doped cuprates it is large at a finite wavevector q, roughly corresponding to 2kF , where kF is the Fermi momentum. The maximum electronic polarization is as large as the ionic polarization, just as in the case of the titanates. This explains the strong e–p coupling of the Cu–O bond-stretching mode for q > 0. Moreover, the phononinduced charge transfer is spin polarized due to the strong electron correlation effect, and therefore the phonon strongly modifies the spin excitation energies and spin correlations [48]. This extra spin–phonon coupling may justify calling the e–p coupling in this system unconventional.

5.4.3 Vibronic Mechanism of High Temperature Superconductivity The earlier discussions above demonstrate the unconventional nature of the e–p coupling for the Cu–O bond-stretching mode in the cuprates, but would it be strong enough to explain the high temperature superconductivity? Within

88

T. Egami

the framework of the Bardeen Cooper and Schrieffer (BCS) mechanism [54] the answer is almost certainly negative. The charge carrier density is low, and the e–p coupling estimated from the ARPES data [33] is not strong enough to explain the value of TC over 100 K. However, if phonons were to contribute to the HTSC, it is not likely to be through the conventional BCS mechanism. The BCS theory assumes the Fermi level of electrons to be far larger than the phonon energy, and thus the Fermi velocity is much greater than the sound velocity, and the Migdal theorem [55] applies. However, the carriers that show a large superconducting gap are around the (π, 0) point, which is a saddle point and has a zero velocity. Furthermore the strong electronic spatial inhomogeneity observed by the STM–STS suggests that the BCS picture of fast electrons interacting with the phonon once may not be appropriate. The possible importance of spatial inhomogeneity was first suggested by Gor’kov [56], and was extensively discussed by Phillips [57–59] and by Bianconi [38, 60]. The idea is that inhomogeneity results in charge confinement that produces jumps in electron dispersion and a high density of states (DOS) at the Fermi level (shape resonance). Phillips further speculated that the resultant superconductivity is highly inhomogeneous, and has filamental character, which seems to be true for underdoped samples. While these arguments assume a conventional e–p coupling with Fermi liquid, it is possible that the e–p coupling is unconventional and nonadiabatic. As a result of confinement the group velocity of electrons is further reduced. If it is comparable to the sound velocity the Migdal theorem is no longer valid, and we have to consider a resonant, nonadiabatic vibronic coupling in real space [61, 62]. Superconductivity due to real space pairs was proposed by Schafroth [63]. Due to mathematical difficulties his theory could not be advanced as much as the BCS theory at that time, but with modern computers it may now be possible to go beyond what has been done nearly a half century ago. Here we briefly sketch the idea of vibronic pairing that could explain the HTSC phenomenon. Coupling of the wave packet of the half-breathing phonon to the electron wave packets, made of regions in k space around four reciprocal lattice points equivalent to the (π, 0) point, could lead to formation of phonon-induced vibronic two-hole bound states with each hole in two different sublattices, thereby reducing the repulsive Coulomb energy and gaining the exchange energy [64]. Strong vibronic coupling of such a bound state can result in large values of TC [62]. The spatial extent of such a bound state is determined by the area of the flat band in the k space near the (π, 0) point, and would form domains of 20–30 ˚ A in size, in agreement with the STM observation. Such a bound state could be the origin of the electronic inhomogeneity. In between the bound states hole-poor regions will develop antiferromagnetic spin correlations, which can create a pseudo-gap. Since the bound state can be formed stably only when the Fermi surface is close to the saddle point of electron dispersion, forming a bridge around the (π, 0) point. In strongly underdoped systems small Fermi surfaces are formed around the (π, π) points

5 Interplay of Spin, Charge, and Lattice in Manganites and Cuprates

89

and the energy at the (π, 0) point is low, which makes the bound state unstable. In the overdoped state a large Fermi surface is formed and the energy at the (π, 0) point is high, again making the bound state unstable. Thus the phonon-induced vibronic bound state can be formed only in the intermediate hole densities when the Fermi level is close to the saddle point, explaining why TC forms a dome as a function of doping [64]. While this scenario needs further development and experimental and theoretical verification, it appears to explain so many properties of this complex system, and is worth pursuing further.

5.5 Conclusions It is clear that the complex interplay of spin, charge, orbital, and lattice degrees of freedom is the underlying mechanism that produces remarkable properties such as HTSC and CMR in transition metal oxides. In trying to disentangle this intertwined state of multiple variables we have to deal with all these factors together on the same footing. Such an approach, however, marks strong departure from the past practice, where we usually start from the state well defined by a dominant interaction, and take into account of less dominant forces by perturbation. In some cases we can start from the balanced state defined by the mean-field approximation and correct for the deviations. However, collective effects of multiple degrees of freedom can render the meanfield approximation invalid, particularly in the close vicinity of the critical point. A novel approach, which appears to be more effective in this new world of complexity, is the local approach, both in experiment and theory. As examples of local approach we discussed two subjects, the polaron stability in the CMR manganites and the vibronic mechanism of high temperature superconductivity. The latter is a mere speculation at this moment, and further research is warranted to test this scenario. When we succeed in constructing a full theory of the vibronic mechanism it could form a new platform for the science of strongly interacting complex systems. The research in this field is important and exciting even for this reason alone. Acknowledgments The author acknowledges his collaborators, in particular, D. Louca, P. Piekarz, M. Tachiki, J.-H. Chung, R.J. McQueeney, M. Arai, M. Yethiraj, H.A. Mook, C. Frost, S. Tajima and Y. Endoh for their contributions, and K.A. M¨ uller, J.B. Goodenough, L. Gor’kov, J.C. Phillips, P.W. Anderson, A. BussmannHolder, N. Nagaosa, M. Khomoto, Z.-X. Shen, A. Lanzara, D. Mihailovic, E. Weger, J.J. Tu, E.W. Plummer, A.R. Bishop, and P. Dai for valuable insights, discussions, and comments. This work was supported by the National Science Foundation through DMR01-02565.

90

T. Egami

References 1. J.G. Bednorz, K.A. M¨ uller: Z. Phys. B 64, 189 (1986) 2. S. Jin, T.H. Tiefel, M. McCormack, R.A. Fastnacht, R. Ramesh, L.H. Chen: Science 264, 413 (1994) 3. G.H. Jonker, J.H. Van Santen: Physica 16, 337 (1959); J.H. Van Santen, G.H. Jonker: Physica 16, 599 (1950) 4. A.J. Millis, P.B. Littlewood, B.I. Shraiman: Phys. Rev. Lett. 74, 5144 (1995) 5. S.J.L. Billinge, R.G. DiFrancesco, G.H. Kwei, J.J. Neumeier, J.D. Thompson: Phys. Rev. Lett. 77, 715 (1996) 6. C.H. Booth, F. Bridges, G.J. Snyder, T.H. Geballe: Phys. Rev. B 54, R15606 (1996) 7. D. Louca, T. Egami, E.L. Brosha, H. R¨ oder, A.R. Bishop: Phys. Rev. B 56, R8475 (1997) 8. D. Louca, T. Egami: Phys. Rev. B 59, 6193 (1999) 9. L. Vasiliu-Doloc, S. Rosenkranz, R. Osborn, S.K. Sinha, J.W. Lynn, J. Mesot, O.H. Seeck, A.J. Fedro, J.F. Mitchell: Phys. Rev. Lett. 83, 4393 (1999) 10. B.J. Campbell, S.K. Sinha, R. Osborn, S. Rosenkranz, L. Vasiliu-Doloc, J.F. Mitchell, Z. Islam, O.H. Seeck, J.W. Lynn: Phys. Rev. B 67, 020409 (2003) 11. T. Egami, S.J.L. Billinge: Underneath the Bragg Peaks: Structural Analysis of Complex Materials (Pergamon, Oxford, 2003) 12. T. Proffen, T. Egami, S.J.L. Billinge, A.K. Cheetham, D. Louca, J.B. Parise: Appl. Phys. A 74, S163 (2002) 13. T. Egami: Z. Kristallogr. 219, 122 (2004) 14. T. Egami, in: Structure and Bonding, ed. by J.B. Goodenough, Vol. 98 (Springer, Berlin Heidelberg New York, 2001), p. 115 15. E. Dagotto, T. Hotta, A. Moreo: Phys. Rep. 344, 1 (2001) 16. H. R¨ oder, J. Zang, A.R. Bishop: Phys. Rev. Lett. 76, 1356 (1996) 17. A.J. Millis, B.I. Shraiman, R. Mueller: Phys. Rev. Lett. 77, 175 (1996) 18. H.Y. Hwang, S.W. Cheong, P.G. Radaelli, M. Marezio, B. Batlogg: Phys. Rev. Lett. 75, 914 (1995) 19. A.P. Ramirez: J. Phys.: Condens. Matter 9, 8171 (1997) 20. P.G. Radaelli, M. Marezio, H.Y. Hwang, S.W. Cheong, B. Batlogg: Phys. Rev. B 54, 8992 (1996) 21. J.A. Fernandez-Baca, P. Dai, H.Y. Hwang, C. Kloc, S.W. Cheong: Phys. Rev. Lett. 80, 4012 (1998) 22. T. Holstein: Ann. Phys. (Paris) 8, 325 (1959) 23. T. Egami, D. Louca: Phys. Rev. B 65, 094422 (2002) 24. D. Louca, T. Egami, W. Dmowski, J.F. Mitchell: Phys. Rev. B 64, R180403 (2001) 25. J.M. Tranquada, B.J. Sternlieb, J.D. Axe, Y. Nakamura, S. Uchida: Nature (London) 375, 561 (1995) 26. M.K. Crawford, R.L. Harlow, E.M. McCarron, W.E. Farneth, J.D. Axe, H. Chou, Q. Huang: Phys. Rev. B 44, 7749 (1991) 27. Y. Maeno, N. Kakei, Y. Tanaka, T. Tomita, F. Nakamura, T. Fujita, in: Lattice Effects in High-TC Superconductors, ed. by Y. Bar-Yam, T. Egami, J.M. de Leon, A.R. Bishop (World Scientific, Singapore, 1992), p. 542 28. J.D. Jorgensen, B. Dabrowski, S. Pei, D.G. Hinks, L. Soderholm, B. Morosin, J.E. Schirber, E.L. Venturini, D.S. Ginley: Phys. Rev. B 38, 11337 (1988)

5 Interplay of Spin, Charge, and Lattice in Manganites and Cuprates

91

29. R.J. Cava, A. Santoro, D.W. Johnson, W.W. Rhodes: Phys. Rev. B 35, 6716 (1987) 30. P.G. Radaelli, D.G. Hinks, A.W. Mitchell, B.A. Hunter, J.L. Wagner, B. Dabrowski, K.G. Vandervoort, H.K. Viswanathan, J.D. Jorgensen: Phys. Rev. B 49, 4163 (1994) 31. R.D. Shannon: Acta. Crystallogr. A 32, 751 (1976) 32. P.W. Anderson: The Theory of Superconductivity in the High-TC Cuprates (Princeton University Press, Princeton, 1997) 33. A. Lanzara, P.V. Bogdanov, X.J. Zhou, S.A. Kellar, D.L. Feng, E.D. Lu, T. Yoshida, H. Eisaki, A. Fujimori, K. Kishio, J.I. Shimoyama, T. Noda, S. Uchida, Z. Hussain, Z.X. Shen: Nature (London) 412, 510 (2001) 34. S.H. Pan, J.P. O’Neal, R.L. Badzey, C. Chamon, H. Ding, J.R. Engelbrecht, Z. Wang, H. Eisaki, S. Uchida, A.K. Gupta, K.W. Ng, E.W. Hudson, K.M. Lang, J.C. Davis: Nature (London) 413, 282 (2001) 35. K.M. Lang, V. Madhavan, J.E. Hoffman, E.W. Hudson, H. Eisaki, S. Uchida, J.C. Davis: Nature (London) 415, 412 (2002) 36. T. Egami, S.J.L. Billinge, in: Physical Properties of High Temperature Superconductors V, ed. by D. Ginsberg (World Scientific, Singapore, 1996), p. 265 37. N.L. Saini, H. Oyanagi, T. Ito, V. Scagnoli, M. Filippi, S. Agrestini, G. Campi, K. Oka, A. Bianconi: Eur. Phys. J. B 75, 36 (2003) 38. A. Bianconi, N.L. Saini, A. Lanzara, M. Missori, T. Rossetti, H. Oyanagi, H. Yamaguchi, K. Oka, T. Ito: Phys. Rev. Lett. 76, 3412 (1996) 39. E.S. Bozin, G.H. Kwei, H. Takagi, S.J.L. Billinge: Phys. Rev. Lett. 84, 5856 (2000) 40. L. Pintschovius, N. Pyka, W. Reichardt, A.Y. Rumiantsev, N.L. Mitrofanov, A.S. Ivanov, G. Collin, P. Bourges: Physica C 185–189, 156 (1991) 41. L. Pintchovius, W. Reichardt, in: Physical Properties of High Temperature Superconductors IV, ed. by D. Ginsberg (World Scientific, Singapore, 1994), p. 295 42. R.J. McQueeney, Y. Petrov, T. Egami, M. Yethiraj, G. Shirane, Y. Endoh: Phys. Rev. Lett. 82, 628 (1999) 43. T. Egami, J.H. Chung, R.J. McQueeney, M. Yethiraj, H.A. Mook, C. Frost, Y. Petrov, F. Dogan, Y. Inamura, M. Arai, S. Tajima, Y. Endoh: Physica B 62, 316 (2002) 44. J.H. Chung, T. Egami, R.J. McQueeney, M. Yethiraj, M. Arai, T. Yokoo, Y. Petrov, H.A. Mook, Y. Endoh, S. Tajima, C. Frost, F. Dogan: Phys. Rev. B 67, 014517 (2003) 45. S. Ishihara, T. Egami, M. Tachiki: Phys. Rev. B 55, 3163 (1997) 46. Y. Petrov, T. Egami: Phys. Rev. B 58, 9485 (1998) 47. W.P. Su, J.R. Schrieffer, A.J. Heeger: Phys. Rev. Lett. 42, 1698 (1979) 48. P. Piekarz, T. Egami: Phys. Rev. B 72, 054530 (2005) 49. D.N. Basov, E.J. Singley, S.V. Dordevic: Phys. Rev. B 65, 054516 (2002) 50. H.J.A. Molegraaf, C. Presura, D. van der Marel, P.H. Kes, M. Li: Science 295, 2239 (2002) 51. A.F. Santander-Syro, R.P.S.M. Lobo, N. Bontemps, Z. Konstantinovic, Z.Z. Li, H. Raffy: Eur. Phys. Lett. 62, 568 (2003) 52. R.D. King-Smith, D. Vanderbilt: Phys. Rev. B 47, 1651 (1993) 53. R. Resta: Rev. Mod. Phys. 66, 899 (1994) 54. J.R. Schrieffer: Theory of Superconductivity (Benjamin, New York, 1964) 55. A.B. Migdal: Sov. Phys. JETP 1, 996 (1958)

92 56. 57. 58. 59. 60. 61. 62. 63. 64.

T. Egami L.P. Gor’kov, A.V. Sokol: JETP Lett. 46, 420 (1987) J.C. Phillips: Phys. Rev. B 41, 8968 (1990) J.C. Phillips: Phys. Rev. Lett. 88, 216401 (2002) J.C. Phillips, A. Saxena, A.R. Bishop: Rep. Prog. Phys. 66, 2111 (2003) A. Bianconi: Int. J. Mod. Phys. B 14, 3289 (2000) J.B. Goodenough, J. Zhou: Phys. Rev. B 42, 4276 (1990) M. Tachiki, M. Machida, T. Egami: Phys. Rev. B 67, 174506 (2003) M.R. Schafroth: Phys. Rev. 111, 1958 (72) T. Egami: J. Phys. Chem. Solids, in press

6 Neutron Scattering Studies of Anomalous Phonon Behavior in Functional Materials S.M. Shapiro

6.1 Introduction Functional materials can be defined as those materials with crucial materials properties that are primarily the output arising from the electrical, magnetic, or mechanical responses to changes in the environment, such as temperature, stress, electrical or magnetic fields. These materials differ from what are called structural materials, which are utilized for their strength and other mechanical properties. In most cases the strong response of the useful functional materials arises from the proximity of operating conditions to a phase transition in these materials, such as metal–insulator (MI) transitions, the onset of ferroelectricity, ferromagnetism, superconductivity or structural transitions such as martensitic phase transitions in shape-memory alloys. These phase transitions, together with their associated functional responses are known to be highly sensitive to the presence of impurities, alloying additions, crystalline defects and their state of organization, i.e., to the atomic scale structures, nanostructures and microstructures. In this chapter the focus will be on the magnetic and structural properties of several functional materials with particular emphasis on characterizing their properties by inelastic neutron scattering. We shall consider the strong interaction between the lattice and the electronic or magnetic properties of the material as revealed by studying the lattice dynamics of the system. First there will be a brief description of the neutron scattering technique and what is measured followed by a discussion of the types of phonon anomalies occurring as a result of electron–phonon coupling. Examples from the literature will be given of these interactions observed in the giant or colossal magnetoresistance materials and high temperature superconductors. The chapter will conclude with a discussion of the anomalous phonon behavior observed in magnetic shape memory alloys.

94

S.M. Shapiro

6.2 Neutron Scattering Neutrons are an ideal probe for studying the lattice and spin dynamics of solids [1]. The wavelength of the neutron is comparable to the inter-atomic spacing in solids, (1–2 ˚ A) and the energies of the neutrons are comparable to lattice energies (∼50 meV). The energy resolution needed in an experiment to measure excitations is on the order of 10%, which is readily achievable in a triple axis spectrometer. With the new generation of synchrotron X-ray sources, inelastic X-ray scattering is demonstrating that it, too, can measure dispersion curves in solids [2]. Since the incident energy of the photons is in the keV regime, an energy resolution of 10−6 is needed to probe the lattice dynamics. Nevertheless, the flux is so great with synchrotron X-ray sources that it is now feasible to do such high-resolution experiments. The advantage with X-rays is that small samples can be studied along with thin films. However, for magnetic scattering the coupling of neutrons is more direct and still the preferred method. In an X-ray or neutron experiment an incident beam with a well defined energy (Ei ) and momentum (ki ) prepared by a monochromator crystal strikes the sample. The particles are scattered by the sample and their energy analyzed by an analyzer crystal. The rules of momentum and energy conservation govern the entire process: Q = τ + q = ki − kf ,
ω = Ei − Ef ,

(6.1) (6.2)

Q and
ω are the momentum and energy, respectively, transferred to the material and they are inferred by measuring differences between the incident and final neutron wavevectors (ki , kf ) and energies (Ei , Ef ). τ is a reciprocal lattice vector and q is the momentum measured within a Brillouin zone. The neutron also has a spin and this is important in measuring the magnetic properties of a material including the magnetic structure and spin wave excitations. In studying lattice vibrations the neutron couples to the atomic displacements through a neutron–nuclear interaction. The intensity of the scattered neutrons is proportional to the scattering function S(Q, ω), which can be written as S(Q, ω) = [n(ω + 1)] χ

(Q, ω) ,

(6.3)

where [n(ω) + 1] is the Bose–Einstein distribution function and for high temperatures where kT 
ω, n(ω) + 1 is approximated by kT /
ω. χ

(Q, ω) is the imaginary part of the dynamical susceptibility and for the case where neutron scattering from the lattice creates or destroys a single phonon one has

6 Neutron Scattering Studies of Anomalous Phonon Behavior

95

Fig. 6.1. Phonon dispersion curve of Ni2 MnGa measured at room temperature. Data from [3]

χ

(Q, ω) =

1 (2π)3  δ(Q − q − τ ) 2 ν0 τ ,q  1 × F (Q)2 [δ(ω − ωqs ) − δ(ω + ωqs )] , ω qs s

F (Q) is the dynamical structure factor given by  bj (Q · ξ j,s ) ei(Q·dj ) e−Wj . F (Q) = √ m j j

(6.4)

(6.5)

Here, bj is the neutron scattering length for atom j with mass mj . ξ j,s is the eigenvector of atom j of the s phonon branch with frequency ωs and e−Wj is the Debye–Waller factor. By measuring the intensity of a phonon in different Brillouin zones with the same q, one can in principle determine the atomic displacements ξ j,s . Equation (6.5) is used to separate out the transverse from longitudinal modes due to the dot product. With a triple axis instrument it is straightforward to keep q constant and vary ω during an experiment. Doing this for several q with different τ a dispersion curve can be measured. Figure 6.1 shows the phonon dispersion curve measured for the ferromagnetic shape memory alloy Ni2 MnGa [3], which will be discussed in more detail later. The most obvious feature is the very low energy of the TA2 mode propagating along the [ζζ0] direction and the strong dip around ζ = 1/3. The arrows indicate anomalies that are most likely due to strong electron–phonon coupling at particular wavevectors predicted by the Fermi surface.

6.3 Phonon Anomalies A lattice dynamical theory involves calculating the eignevalues of a dynamical matrix, which is proportional to the interatomic force constants [4]. A simple

96

S.M. Shapiro

Born–von K´ arm´an model is usually a good starting point. This essentially is a fit of a dispersion curve to a number of Fourier components. For a nearly harmonic solid, the acoustic dispersion curves are sinusoidal and there are only a few Fourier components needed which are proportional to near or next-near neighbor force constants. If there are anomalies as shown in Fig. 6.1, more terms are needed in the Fourier analysis and the simple Fourier expansion is no longer adequate and more sophisticated methods are needed. In this section we shall consider here two types of phonon anomalies: (1) Kohn anomalies which are directly related to the Fermi surface and very often are subtle features or kinks in a dispersion curve; and (2) soft mode anomalies that are considered as precursor to a phase transformation and show large temperature or pressure dependence of the phonon energies at a particular wavevector. 6.3.1 Kohn Anomalies Kohn anomalies are a direct consequence of electron–phonon coupling. We can describe the dynamical matrix as a sum of two terms [5] D = DSR + DLR .

(6.6)

DSR are the short range mostly harmonic terms that can be described by a Born–von K´ arm´an analysis. In a metal DLR are longer range terms such as electron–phonon interactions. Many authors calculated the electron–phonon coupling for metals, but the approach of Varma and Weber [6] is generally used. This term is related to the generalized electronic susceptibility and is a negative quantity, which would reduce the eigenvalues of D and thus cause a reduction in the phonon frequency  fk,µ (1 − fk+q,ν ) Mk,µ;k+q,ν Mk+q,ν;k,µ . (6.7) DLR ≈ − εk+q,ν − εk,µ k,µ,ν

Here, f is the Fermi function and εk,µ are energies of the µth electronic band with momentum k. It can be seen that if there is a wavevector spanning the Fermi surface the denominator will be small and DLR can be large enough to cause a reduction in the phonon energy. The quantities, M, are the complicated matrix elements giving the strength of the electron–phonon coupling. Both quantities, the electronic structure and the matrix elements, are amenable to calculation and determine the strength of the observed anomalies. Figure 6.2 demonstrates the validity of the calculations [7]. The data points were measured by inelastic neutron scattering after the calculations of Zhao and Harmon [8], who predicted anomalies in the binary alloy, Ni50 Al50 at the positions of the arrows. As it can be seen they are in close agreement with the experimental results, which were obtained after the predictions were made. Not only were the positions of the anomalies well predicted, but their strengths as well. It is stronger along the [ζζζ] direction than the [ζ00] direction. The crosses in Fig. 6.2 are from earlier measurements on this material [9].

6 Neutron Scattering Studies of Anomalous Phonon Behavior

97

Fig. 6.2. The measured phonon dispersion curve of the [ζζζ]–TA (top) and the [ζ00]–LA (bottom) branch in the Ni50 Al50 alloy. The arrows indicated the predicted q-values of the anomaly by Zhao and Harmon [8]. Data from [7]

These authors measured the dispersion curve at well-separated q-values and were not able to confirm the anomalies predicted later. It shows the importance of having sufficient number of data points to accurately determine the shape of the dispersion curve. 6.3.2 Soft Modes Soft modes in structural phase transitions are another example of phonon anomalies, but that are more dramatic than the small features of Kohn anomalies. The soft mode theory of structural phase transitions was first proposed by Anderson [10] and Cochran [11] to describe the phase transition occurring in ferroelectrics. In this theory there is a particular lattice vibration whose displacements are those that are needed to transform the system from its high temperature-high symmetry phase to its lower symmetry-low temperature ground state. The restoring forces for this mode get weaker and weaker as the transition is approached and the frequency tends to zero. Looking at

98

S.M. Shapiro

the dynamical matrix of (6.6) the short-range forces can be viewed as harmonic forces which are temperature independent. They can be parametized as DSR = −KTc . The longer ranged anharmonic forces are proportional to temperature: DLR = KT . The eigenvalue of the dynamical matrix is proportional to the square of the mode energy,
ω (
ω)2 = K(T − Tc ) .

(6.8)

This shows that the phonon energy would go to zero at the critical temperature, Tc . The crystal becomes unstable at this temperature and transforms into its low temperature structure. The soft mode theory of phase transitions has been demonstrated in many systems [12]. The prototypical one is SrTiO3 , which undergoes an antiferroelectric transition [13] near 100 K. The soft mode is a zone boundary phonon and the temperature dependence is shown in Fig. 6.3. The bottom portion shows the energy of the zone boundary mode measured at the (1/2, 1/2, 1/2)

Fig. 6.3. The temperature dependence of the soft mode measured at the (1/2, 1/2, 1/2) zone boundary (bottom) and the intensity of the elastic scattering at this position (top), which becomes a new Bragg peak in the low temperature phase. Data from [13]

6 Neutron Scattering Studies of Anomalous Phonon Behavior

99

R-point of the Brillouin zone. The solid line is a fit to (6.8) with Tc = 108 K. The upper portion shows the intensity of the elastic scattering as a function of temperature and reveals a new Bragg peak appearing below the transition temperature corresponding to a lowering of the symmetry.

6.4 Phonon Anomalies in the Manganites The ferromagnetic perovskites of the type La1−x Mex MnO3 (Me=Ca or Sr) are under intense study due to the important discovery of their giant magnetoresistance for certain regions of doping [14]. In these regions the ground state is ferromagnetic and at temperatures above the Curie temperature they become insulating. The magnetic ordering can partly be explained by the double exchange mechanism. However, to explain the temperature dependent behavior of the resistivity requires a strong electron–phonon interaction arising from the Jahn–Teller splitting of the Mn3+ d-band eg levels. The structural and magnetic phase diagrams are complex and exhibit regions where charge ordering of the Mn3+ and Mn4+ occurs. The exciting physics in these systems lies in the coupling between the spin, the charge, the electrons, the lattice, and orbital degrees of freedom – all the fundamental entities in solid state physics. Below we review one of the experiments on this system where inelastic neutron scattering experiments have shed light on the interaction between the spins and the lattice. The lattice vibrations associated with the Jahn–Teller distortions involve the atomic motions of the MnO6 octahedra and should show anomalous behavior as the temperature is reduced below the MI phase transition. Zhang et al. [15] studied this in detail in La0.7 Ca0.3 MnO3 above and below the ferro-to-paramagnetic transition temperature, Tc ∼ 240 K, which coincides with the MI transition. Figure 6.4 shows the dispersion curve of the optic modes along the [ζ00] direction measured at 10 K. The interesting mode is the highest energy mode with energy
ω3 ∼ 71 meV at ζ = 0. This is the in plane MnO3 bond stretching mode, which is the Jahn–Teller mode. Calculations and measurements in the undoped materials show that this upper mode is essentially dispersionless or should bend upwards as ζ increases. Instead, the mode bends downward with a negative dispersion. This anomalous effect is related to the charge ordering that occurs at the MI phase transition. Figure 6.5 shows the temperature dependence of this mode. It exhibits anomalous damping and a decrease of intensity throughout the Brillouin zone with increasing temperature. It is argued that this anomalous behavior is due to the decoherence effects of the local dynamic phase fluctuations associated with short-range polaron or charge/orbital orderings when the system approaches Tc .

100

S.M. Shapiro

Fig. 6.4. Measured optical phonon dispersion curves in La0.7 Ca0.3 MnO3 at 10 K. The open circles were measured in the Sr doped material. The solid lines are the results calculated by a screened shell model under cubic symmetry. Data from [15]

6.5 Phonon Anomalies in High Temperature Superconductors Since the discovery of high temperature superconductors nearly 20 years ago, the basic mechanism behind the electron pairing leading to the superconductivity remains unexplained [16]. All of the cuprates have interesting magnetic properties that change dramatically with doping. Therefore, it is thought that magnetism plays an important role in the coupling mechanisms that lead to the high Tc . Up until recently electron–phonon coupling was thought to play a minor role in establishing the high superconducting transition temperatures. The discovery of stripes [17] in the cuprates – the separation of spin and charge – has given more impetus to the study of phonon anomalies in these compounds. It is unclear whether the appearance of stripes helps or hinders the onset of superconductivity. Studies of phonon behavior can help elucidate the role of charge ordering in superconductivity because phonons with a displacement pattern closely related to the charge stripe wavevector should have anomalously low energies analogous to what has been observed as precursor phenomenon to charge-density wave order in low-dimensional metals [18]. The phonon most likely connected to stripes is the highest frequency phonon, a longitudinal optic branch involving Cu–O stretching vibrations. A study of the phonon anomalies in the high temperature superconductor YBCO (YBa2 Cu3 O6.6 ) with Tc ∼ 60 K was performed by Pintschovius

6 Neutron Scattering Studies of Anomalous Phonon Behavior

101

Fig. 6.5. Temperature dependence of energy scans at the Brillouin zone center in La0.7 Ca0.3 MnO3 . ω3 is the Mn–O bond stretching Jahn–Teller mode. Data from [15]

et al. [19]. They chose this material because incommensurate magnetic fluctuations are well documented and the wavevector of any phonon anomalies would be twice the wavevector of the magnetic fluctuations if the anomaly is related to the stripes. The structure of this crystal is orthorhombic and the crystal is twinned in the (0, 0, 1) plane. It is difficult, therefore, to separate the modes propagating along the [ζ00] and [0ζ0] directions, but Pintschovius et al. [19] were able to do so. Figure 6.6a shows the calculated dispersion curves of the ∆1 longitudinal optic modes calculated from an interaction potential model. The red and blue color corresponds to modes propagating along the [ζ00] and [0ζ0] directions, respectively. The phonons propagating along [ζ00] have a slightly higher energy because of the smaller Cu–O distance. The upper branches between 65 and 75 meV are the Cu–O stretching vibrations. In Fig. 6.6b the measured phonon dispersion curves with the same color representation are shown. The most anomalous feature is the rapid drop of the frequencies for the high-energy phonons propagating along the [0ζ0] direction at about halfway to the zone boundary at ζ ∼ 0.25.

102

S.M. Shapiro

Fig. 6.6. Dispersion curves of the high energy longitudinal phonons in YBCO0.66 . (a) Calculated curves from an interaction potential mode. The red (blue) curve corresponds to the [ζ00] ([0ζ0]) direction. (b) Measured dispersion curves. Red (blue) correspond to the [ζ00] ([0ζ0]) direction. The black symbols denote phonon peaks where the assignment could not be resolved. From [19]

If one studies the displacement pattern associated with this phonon it is seen that it is readily compatible with dynamic charge stripe formation with a period of about 4 times the b lattice parameter. However, there are some inconsistencies when compared to the magnetic data, which suggests that the modulation is along the a direction and the above phonon measurements imply that the charge modulation is along the b direction. This was discussed in a later paper where the phonons in an optimally doped YBCO6.95 sample were studied [20]. In this experiment they studied the dispersion and the temperature dependence of ∆4 phonons. These phonons also involve Cu–O stretching, but they differ from the ∆1 symmetry in that the elongations in the Cu–O bi-layer are out of phase for ∆4 and in phase for ∆1 . The behavior of the two are similar, but the interactions of the ∆1 modes with c-axis polarized branches leads to complications in the measurements. These latest results provide more convincing arguments that the phonon anomalies in the Cu–O stretching mode indicate a dynamic one-dimensional charge density modulation with the Cu–O planes.

6.6 Ferromagnetic Shape Memory Alloys Most shape memory materials are tuned by varying the temperature. However, there is a growing list of shape memory alloys that, in addition to the

6 Neutron Scattering Studies of Anomalous Phonon Behavior

103

Martensitic transformation exhibit ferromagnetic ordering. The major interest in these materials arises from the possibility of tuning the strain and the shape memory transition by applying a magnetic field, rather than temperature. This opens up a new domain for actuator applications since changing a magnetic field is a quicker operation than changing temperature. In Ferromagnetic Shape Memory Alloys (FSMA), the Curie temperature (TC ) is weakly dependent on composition and the Martensitic temperature (TM ) is strongly dependent upon the composition. TM can therefore be tuned to be higher or lower than TC . The fundamental excitations in the spin system are magnons and those of the lattice are the phonons. Since these collective excitations are proportional to the couplings between the individual spins (magnons) and atoms (phonons) a study of them is important to understanding the fundamental mechanism of the transformations and the coupling between the spin and the lattice. Precursor effects such as critical scattering and softening of magnon energies are well known in magnetic transitions, but not ubiquitous in the Martensitic transformations. A study of the temperature dependence of the magnons and phonons is therefore important in delineating the precursor phenomena and the interaction of the magnetic and structural properties. 6.6.1 Iron-Based Alloys The coupling of the structural and magnetic properties in iron-based alloys has been known for many decades. Invar alloys have the interesting and technologically very important property of having zero or slightly negative thermal expansion over a very wide temperature range [21]. This useful property has been exploited in the construction of high precision mechanical instruments, but is not fully explained. It disappears above the Curie temperature so is related to the magnetic ordering. Most theories are based upon the volume instabilities of the magnetic moments that give rise to a magneto-elastic coupling that will offset the normal lattice expansion. Understanding the temperature dependent behavior of the phonons will shed light on this coupling. Fe65 Ni35 is the canonical invar alloy but other Fe-alloys, such as Fe–Pt and Fe–Pd exhibit similar invar properties. All of these show a change in phonon dispersion as the system orders ferromagnetically. Figure 6.7 shows the dispersion curve [22] of the [ζζ0] TA2 phonon branch for several invar alloys along with bcc Fe. At temperatures above the Curie temperature the dispersion follows the normal linear acoustic behavior for small ζ-values. For temperatures below the Curie temperature an anomalous curvature in the dispersion over nearly the entire Brillouin zone develops. A more detailed picture is shown for Fe72 Pd28 in Fig. 6.8 [23]. At 676 K, which is higher than the Curie temperature TC ∼ 575 K, the dispersion shows the usual linear behavior. Below TC it develops a concave behavior, which increases as the temperature is lowered. This is demonstrated in Fig. 6.9 where the ratio of the phonon energies measured below TC to that above TC is plotted for several ζ-values as a function of temperature. For all ζ, the softening starts at TC and is

104

S.M. Shapiro

Fig. 6.7. [ζζ0]-TA2 dispersion curves for several Fe-based alloys and bcc Fe. Data from [22]

largest for the smaller ζ. This behavior is also observed in the other alloys shown in Fig. 6.7 and demonstrates the coupling between the magnetic ordering and the phonon behavior. A measure of the phonon linewidth also shows an increase with decreasing temperature starting at the Curie temperature as shown in Fig. 6.10. This broadening was confined only to the [110] TA2 branch which exhibits the softening. The other branches measured along symmetry directions exhibit normal behavior. 6.6.2 Heusler-Based Alloys Ni2 MnGa This material has generated a tremendous amount of interest over the past 10 years largely due to the coexistence of a shape memory Martensitic transition with ferromagnetic ordering and the possibility of inducing a large strain by application of a magnetic field. Recently it was shown that a 10% strain can be induced by application of a magnetic field at room temperature [24]. This is an enormous amount; much larger than piezoelectric induced strain. The symmetry of the high temperature phase of this material is cubic with the fcc Heusler-L21 structure (Fm3m) with Curie temperature near 380 K, which

6 Neutron Scattering Studies of Anomalous Phonon Behavior

105

Fig. 6.8. Temperature dependence of the [ζζ0]-TA2 branch for Fe72 Pd28 . Data from [23]

Fig. 6.9. Temperature dependence of the [ζζ0]-TA2 phonons for Fe72 Pd28 , normalized by the phonon energy above TC . Data from [23]

varies little with composition. The Martensitic transformation temperature varies drastically with composition and the low symmetry phase is either orthorhombic or tetragonal, depending upon the composition. Figure 6.11 shows the phase diagram of Ni2 MnGa constructed from published reports [25]. The quantity α is a weighted composition of Mn and Ga, which is easily converted

106

S.M. Shapiro

Fig. 6.10. Temperature dependence of the phonon linewidth of the [ζζ0]-TA2 branch for Fe72 Pd28 . Data from [23]

Fig. 6.11. Phase diagram of Ni2 MnGa from [25]. The filled circles and filled triangles are the Martensitic and Curie temperatures, respectively. The open triangles represent the transition to the intermediate modulated phase. The hatched region in the inset represents the composition range from which the data have been taken

to the electron per atom ratio. For α = 25 the composition is stoichiometric and e/a = 7.5. The solid circles show the variation of the Martensitic transformation temperature and the solid triangles represent the Curie temperature, which is nearly independent of composition, whereas the Martensitic temperature varies dramatically with composition. The open triangles delineate a phase transformation into an intermediate modulated cubic phase, which exists just prior to the transformation to the Martensitic phase.

6 Neutron Scattering Studies of Anomalous Phonon Behavior

107

Fig. 6.12. Temperature dependence of anomalous TA2 branch in Ni2 MnGa. Data from [26]

The phonon dispersion curve for Ni2 MnGa is shown in Fig. 6.1. The obvious anomaly is the [ζζ0]-TA2 mode. A more detailed temperature study of the low energy portion of this branch is shown in Fig. 6.12 [26]. At the highest temperature, 400 K, which is above the Curie temperature, the dispersion branch shows a small wiggle. Since the anomaly is still present about the Curie temperature this suggests that the coupling between the magnetic and lattice properties is small. As the temperature is lowered towards the Martensitic temperature, TM ∼ 220 K, a distinct minimum develops at ζ0 = 0.33. The energy of the mode never goes to zero and reaches a minimum at TI = 260 K, 40 K above TM and then begins to increase again, still within the cubic phase. This suggests that another phase develops between TI and TM and coexists with the parent phase. This is the ferromagnetic intermediate phase shown in the phase diagram of Fig. 6.11. The elastic scans along the [ζζ0] transverse direction are shown in Fig. 6.13. At elevated temperature a small diffuse bump is observed at ζ0 = 0.33, the same wavevector as the anomaly seen in the dispersion curve. As T decreases the intensity grows as the frequency decreases. Around 260 K there is a strong increase in the intensity and it becomes narrower and Bragg-like, indicating that an ordered modulation of the cubic phase sets in. This temperature is the same as the minimum of the dispersion curve. This type of elastic scattering associated with a phonon softening is ubiquitous in many structural transformations and is related to defects of the lattice [27]. The phonon softening is usually viewed as a precursor to the low temperature phase. It is worth noting that in the Martensitic phase, there is a modulation of the lattice [28] with a wave vector ζ0 ∼ 0.43, which is quite different from 0.33. Thus the phonon

108

S.M. Shapiro

Fig. 6.13. Temperature dependence of the elastic scattering along the [ζζ0] transverse direction. Data from [26]

softening should not necessarily be viewed as a precursor to the Martensitic phase, but more appropriately a precursor to the intermediate phase. Since the phonon dispersion curve did not become linear above the Curie temperature it was suggestive that the coupling to the magnetization is weak and there was little effect of the magnetic ordering on the phonon behavior. This is not strictly correct. When the square of the phonon frequency measured at ζ0 ∼ 0.33 is plotted as a function of temperature a change of slope occurs at the magnetic ordering temperature [28] as seen in Fig. 6.14. What is the origin of the phonon softening and modulation observed at the wavevector ζ = 0.33? Self-consistent first principles calculations of the electron energy spectrum, density of electronic states, and the generalized susceptibility and of various Fermi surface cross sections have been carried out by several groups. One group [29] concluded that Fermi surface nesting could not be the driving force of the phonon softening because the calculation showed that the nesting wavevector in the ferromagnetic phase of Ni2 MnGa is ζ0 = 0.42, which is substantially larger than the experimental value of ζ0 = 0.33 as shown above. This was reconciled by a later study [30] where the changes in the Fermi surface upon magnetic order were calculated. Since the phonon softening persists down to TM , which is well above zero, the magnetization is not fully saturated. Lee et al. [30] showed that the nesting vector does, indeed, depend upon the magnetization. They found that the Fermi surface nesting is optimized at 80% of full magnetization and the nesting vector is ζ0 = 1/3, the same wavevector as the observed phonon softening. However,

6 Neutron Scattering Studies of Anomalous Phonon Behavior

109

Fig. 6.14. Square of phonon energy of [ζζ0]-TA2 mode, ζ = 0.33 vs. temperature for Ni2.06 Mn0.94 Ga. The Curie temperature for this alloy is TC = 364 K and TM = 284 K. Data from [28]

these calculations would suggest that the nesting vector and therefore the wavevector of the anomaly and its associated diffuse elastic peak should be temperature dependent, which it is not. The 80% magnetization level occurs close to the temperature region near where the premartensitic transition occurs. At lower temperatures in the Martensite phase, the magnetism is near saturation and the nesting vector has shifted to ζ0 = 0.42, consistent with the observations [31]. Studies of single domain samples of Ni2 MnGa in the tetragonal Martensite phase show that the lattice modulation is at ζ0 = 0.42, which corresponds to the weakly temperature dependent phonon anomaly of the [ζζ0] TA2 phonons [31]. Co2 NiGa This compound was recently discovered [32] and is isomorphic to Ni2 MnGa, although much less is known about the material. For example, in the Nicompound, most of the magnetization is localized on the Mn sites, with a small moment (10%) on the nickel sites. The magnetic properties for the Co compound are unknown. It orders around 380 K and the Martensitic transformation temperature is near room temperature. The stoichiometery can be varied and a complete phase diagram has not been mapped out. The variation of the transition temperature with e/a ratio [33] is nearly identical to Ni2 MnGa. A single crystal of Co2 NiGa was grown with composition Co1.92 Ni0.88 Ga1.2 with a TM on cooling near ∼280 K. The phonon dispersion curve and the elastic scattering along the [ζζ0]-TA2 direction were measured

110

S.M. Shapiro

Fig. 6.15. The [ζζ0]-TA2 phonon branch measured in Co1.92 Ni0.88 Ga1.2 at several temperatures within the cubic phase. The dashed line shows the same branch measured in Ni2 MnGa. Data from [34]

in the Austenite phase at several temperatures [34] and is shown in Fig. 6.15. There is no anomaly in the dispersion curve nor is there any temperature dependence. The elastic scans revealed no diffuse peaks and thus no modulation. Therefore, no intermediate phase exists between the Austenite and the cubic phase. The absence of any anomaly and diffuse scattering in Co1.92 Ni88 Ga1.2 is surprising since it is so similar to Ni2 MnGa in terms of structural and magnetic properties. On the other hand, the absence of the intermediate phase and no phonon softening implies that the phonon softening observed in Ni2 MnGa is more likely a precursor to the intermediate phase rather than the Martensite phase. This argument can be extended to the phonon softening observed in other shape memory alloy systems and raises the question of whether phonon anomalies are precursors to the Martensitic transformation or to the premartensitic phase.

6.7 Summary The technologically important functional materials exhibit phase transitions that are driven by a coupling of fundamental quantities, such as electron phonon coupling or a coupling of two different degrees of freedom such as magnetic spins or atomic displacements. It was demonstrated in this chapter that by inelastic neutron scattering studies of the lattice dynamics, a great deal of information about the coupling can be learned. These measurements, coupled to the theory, can lead to an understanding of the driving mechanism

6 Neutron Scattering Studies of Anomalous Phonon Behavior

111

of the transitions. Examples from studies of a diverse group of materials ranging from high Tc compounds, the manganites and magnetic shape memory alloys show anomalies in various relevant phonon branches. Acknowledgments I would like to acknowledge the collaborations with an excellent group of scientists: P. Vorderwisch (Hahn Meitner Institute), A. Zheludev (ORNL), Jason Gardner (BNL/NIST), Guongyong Gu (BNL). Particular thanks go to my colleagues at Ames Laboratory, T. Lograsso and D. Schlagel who grew the excellent crystals for the neutron studies. Work at Brookhaven is supported by the Office of Science, U.S. Department of Energy, under Contract No. DE-AC02-98CH10886.

References 1. G. Shirane, S.M. Shapiro, J.M. Tranquada: Neutron Scattering with a TripleAxis Spectrometer-Basic Techniques (Cambridge University Press, 2002) 2. J. Wong, M. Krisch, D.L. Farber, F. Occelli, A.J. Schwartz, T.C. Chiang, M. Wall, C. Boro, R. Xu: Science 301, 1078 (2003) 3. A. Zheludev, S.M. Shapiro, P. Wochner, L.E. Tanner: Phys. Rev. B 54, 15 045 (1996) 4. M. Born, K. Huang: Dynamical Theory of Crystal Lattices (Oxford University Press, 1954) 5. C.M. Varma, W. Weber: Phys. Rev. Lett. 39, 1094 (1977) 6. C.M. Varma, W. Weber: Phys. Rev. B 19, 6142 (1979) 7. H. Chou, S.M. Shapiro: Phys. Rev. B 48, 16 088 (1993) 8. G.L. Zhao, B.N. Harmon: Phys. Rev. B 45, 2818 (1992) 9. M. Mostoller, R.M. Nicklow, D.M. Zehner, S. Lui, J.M. Mundenar, E.W. Plummer: Phys. Rev. B 40, 2856 (1989) 10. P.W. Anderson: Qualitative considerations on the statistics of the phase transition in batio3 -type ferroelectrics. In: Fizika Dielekrikov, ed. by G.I. Skanavi (Academia Nuak, Moscow, USSR, 1960) pp. 290–295 11. W. Cochran: Adv. Phys. 9, 387 (1960) 12. A.D. Bruce, R.A. Cowley: Structural Phase Transitions (Taylor & Francis, London, 1981) 13. G. Shirane, Y. Yamada: Phys. Rev. 177, 858 (1969) 14. A.M. Goldman: Science 274, 1630 (1996) 15. J. Zhang, P. Dai, J.A. Fernandez-Baca, E.W. Plummer, Y. Tomioka, Y. Tokura: Phys. Rev. Lett. 86, 3823 (2001) 16. P.W. Anderson: The Theory of Superconductiviy in High-Tc Cuprates (Princeton University Press, 1997) 17. J.M. Tranquada, B.J. Sternlieb, J.D. Axe, Y. Nakamura, S. Uchida: Nature 375, 561 (1995) 18. D.E. Moncton, J.D. Axe, F.J. DiSalvo: Phys. Rev. B 16, 801 (1977) 19. L. Pintschovius, W. Reichardt, M. Kl¨ aser, T. Wolf, H. v. L¨ ohneysen: Phys. Rev. Lett. 89, 037 001-1 (2002)

112

S.M. Shapiro

20. L. Pintschovius, D. Reznik, W. Reichardt, Y. Endoh, H. Hiraka, J.M. Tranquada, H. Uchiyama, T. Masui, S. Tajima: cond/mat 0308357 (unpublished); Phys. Rev. B 69, 214506-1 (2004) 21. E.F. Wassermann: The Invar and anti-Invar Effect: finally understood after 100 years? In: The Invar Effect: Centennial Symposium, ed. by J. Wittenauer (Minerals Metals & Materials Society, 1997) p. 51 22. J. K¨ astner, W. Petry, S.M. Shapiro, A. Zheludev, J. Neuhaus, T. Roessel, E.F. Wassermann, H. Bach: Eur. Phys. J. B 10, 641 (1999) 23. M. Sato, B.H. Grier, S.M. Shapiro, H. Miyajima: J. Phys. F: Met. Phys. 12, 2117 (1982) 24. A. Sozinov, A.A. Likhachev, N. Lanska, K. Ullakko, V.K. Lindroos: J. Phys. IV France 112, 955 (2003) 25. A. Gonz` alez-Comas, E. Obrad´ o, L. Ma˜ nosa, A. Planes, V.A. Chernenko, B.J. Hattink, A. Labarta: Phys. Rev. B 69, 7085 (1999) 26. A. Zheludev, S.M. Shapiro, P. Wochner, L.E. Tanner: Phys. Rev. B 51, 11 310 (1995) 27. J.B. Hastings, S.M. Shapiro, B.C. Frazer: Phys. Rev. Lett. 40, 237 (1978) 28. U. Stuhr, P. Vorderwisch, V.V. Kokorin, P.A. Lindg˚ ard: Phys. Rev. B 56, 14 360 (1997) 29. O.I. Velikokhatnyi, I.I. Naumov: Physics of the Solid State 41, 617 (1999) 30. Y. Lee, J. Rhee, B.N. Harmon: Phys. Rev. B 66, 054 424-1 (2002) 31. P. Vorderwisch et al. (unpublished) 32. M. Wuttig, J. Li, C. Craciunescu: Scripta Mater. 44, 2393 (2001) 33. C. Craciunescu, Y. Kishi, T.A. Lograsso, M. Wuttig: Scripta Mater. 47, 285 (2002) 34. S.M. Shapiro et al. (unpublished)

7 The Structures and Transformation Mechanism in the Ferromagnetic Shape Memory Alloy Ni2 MnGa P.J. Brown, T. Kanomata, and M. Matsumoto, K.-U. Neumann, and K.R.A. Ziebeck

7.1 Introduction The phenomenon of coupled magnetic and structural phase transitions is very rare in condensed matter physics. Such systems can exhibit many interesting properties including giant magneto-caloric, magneto-resistance and magnetostriction, and consequently have great technological potential. Of particular interest are materials that can be formed at one temperature TF , then cooled to a lower temperature TM and plastically deformed and on heating to TF regain their original shape. In recovering their shape the alloys can produce a displacement or a force, or a combination of the two. Such behaviour is known as the shape memory effect and usually takes place by change of temperature or by applied stress. However for many applications such as for actuators the transformation is not sufficiently rapid. Poor energy conversion also limits the applicability of many shape memory alloys. Therefore considerable effort has been made to find a magnetic system in which the phase transition can be controlled by a magnetic field at constant temperature. Numerous materials have been studied but the Ni–Mn–Ga system has proved the most encouraging [1]. Since Ni2 MnGa orders ferromagnetically below TC = 365 K the possibility of producing giant field induced strains which are an order of magnitude larger than those observed in rare-earth transition metal alloys has stimulated a large number of investigations aimed at applications [2, 3]. The origin of the shape memory effect in Ni2 MnGa is the thermo-elastic phase change from the cubic L21 Heusler structure to a phase of lower symmetry which takes place at Tm ∼ = 200 K on cooling [4]. Different transformation temperatures and structures [5, 6] have been reported for the low temperature phase. The variations are probably due to non-stoichiometry or stress.

7.2 The Crystal Structure of the Cubic Austenite Phase Intermetallic compounds formed at the stoichiometric composition X2 YZ and with the cubic L21 structure as shown in Fig. 7.1 are usually classified as

114

P.J. Brown et al.

Site L21 B2

A Mn Mn/Ga

B Ni Ni

C Ga Mn/Ga

D Ni Ni

Fig. 7.1. The Heusler L21 and B2 structures

Heusler alloys. The X and Y constituents are normally transition metals and Z a group B element. Initial interest focussed on the ferromagnetic properties of Cu2 MnAl [7] at a time when the antiferromagnetic properties of manganese were not known. Subsequently a large number of Heusler alloys have been discovered with a wide range of physical properties [8]. Particular interest has focussed on X2 MnZ compounds in which the magnetic moment√is confined to the manganese atoms. Since the Mn atoms are separated by a/ 2 with a ∼ 6 ˚ A, direct overlap of the 3d functions is negligible. Consequently these compounds provide good model systems for studying localised magnetic behaviour in metallic systems. Galvanomagnetic measurements indicate that the transport properties are due only to the s and p electrons. Extensive measurements have shown that the physical properties are sensitive to the degree of stoichiometry and the degree of atomic order. However in many studies these parameters have not been established and therefore it is impossible to correlate results from different investigations. Experiments on Ni–Mn–Ga alloys show that stoichiometry strongly influences the transition temperature TM but that the change in Curie temperature is less pronounced. Furthermore, although a wide range of heat treatments is used to make the alloys, a detailed structural analysis is seldom carried out and hence the degree of atomic order remains uncertain. It is frequently assumed that the austentite phase has the L21 structure, but sometimes this is just inferred from the lattice parameter or from observation of fundamental reflections which are order independent. Thus there are very few examples where the correlation between changes in physical properties and the crystallographic structure, which is essential for a complete understanding of shape memory, have been made. Since the atomic numbers of Ni, Mn and Ga (28, 25 and 31) are close it is difficult to obtain a reliable estimate of the atomic order from

7 The Structures and Transformation Mechanism in Ni2 MnGa

115

X-ray powder diffraction measurements. Neutron diffraction which probes the bulk structure (10−2 m), enables a quantitative structural refinement to be obtained from powder data particularly if the profile refinement technique [9] is employed. The nuclear coherent scattering amplitudes of 1.03, −0.373 and 0.723 [10−12 cm] for Ni, Mn and Ga respectively are significantly different making neutron diffraction particularly appropriate for these materials. The Heusler unit cell shown in Fig. 7.1 is comprised of four interpenetrating fcc lattices A, B, C and D with origins at (000), ( 14 14 14 ), ( 12 12 12 ) and ( 34 34 34 ). Bragg reflections are permitted when the Miller indices are unmixed which gives rise to three types of structure factors: • h, k, l all odd F (111) = 4[(fA − fC ) + i(fB − fD )] ;

(7.1)

• h, k, l all even and h + k + l = 4n + 2 F (200) = 4|fA − fB + fC − fD | ;

(7.2)

• h, k, l all even and h + k + l = 4n F (220) = 4|fA + fB + fC + fD | ;

(7.3)

where fA , fB , fC and fD are the average scattering factors of the atoms in the respective sublattices. The reflections for which h + k + l = 4n are the order independent principal reflections. If the alloys are ordered in the L21 structure the structure factors become F (111) = 4|fA − fC | ,

(7.4)

F (200) = 4|fA + fC − 2fB/D | ,

(7.5)

F (220) = 4|fA + fC + 2fB/D | .

(7.6)

A comparison between the square of the structure factors of Ni2 MnGa for both X-rays and neutrons in Table 7.1 shows that the latter technique is more appropriate for structural analysis. Possible types of atomic disorder and models describing disorder have been extensively discussed [9]. Although there is an infinite number of ways in which the atoms XYZ can be distributed over the lattices ABCD there are several types of preferential disorder which frequently occur in Heusler alloys. B2 disorder implies the interchange of atoms

Table 7.1. X-ray and neutron structure factors of Ni2 MnGa in the L21 structure

X-ray (10−24 cm2 ) Neutron (10−24 cm2 )

F 2 (111)

F 2 (200)

F 2 (220)

45.74 19.22

0 46.79

15937.0 92.93

116

P.J. Brown et al.

Fig. 7.2. The effect of preferential disorder on the neutron structure factors of Ni2 MnGa

on the A and C lattices which can lead to a substantial change in properties since the atomic separation between some pairs of Mn atoms is reduced to a/2. The effects of preferential disorder on the structure factors can be characterised using a single parameter α defined as the fraction of either Y or Z atoms not on their ‘correct’ sites. The effects of the various types of disorder on the squares of the structure factors and hence Bragg intensities for neutron scattering are shown in Fig. 7.2.

7.3 Bulk Magnetic Properties Between 200 K and the Curie temperature 370 K the thermal variation of the magnetisation of stoichiometric Ni2 MnGa follows that expected for an isotropic ferromagnet. The structural phase transition at 200 K introduces a preferred axis of magnetisation (c axis) which in low fields ( 0, then ∗ ∗ there is a limited  range of volumefractions, given precisely by [0, λ ] ∪ (1 − λ , 1],

4

where λ∗ =

1 2

1−



1 − 4(µ/|a|2 ) and µ = tr U21 − det U21 − 2, for which there are

austenite/martensite interfaces.

9 Materials with “Unlikely” Properties

169

Fig. 9.2. Illustration of the continuous variation of volume fraction of martensite possible under the cofactor conditions. Distortions accurately drawn using the distortion matrix and twin system given in (9.7)ff

Note the remarkable degree of flexibility indicated by these pictures. In fact, as the volume fraction of martensite goes from 0 to 1, the (perpendicular) directions of maximum and minimum principle strain in the martensite actually exchange places. Thus, these special conditions on lattice parameters do not mean that the martensite is compatible with the austenite because it does not deform with volume fraction changes. On the contrary, it undergoes large average deformations while remaining compatible with austenite. One possible objection to the usefulness of the cofactor conditions is that they only appear to apply to one twin system. As we shall see later, this is not the case. In cubic to monoclinic transformations, if the cofactor conditions are satisfied for one twin system, then they are satisfied for many twin systems.

9.6 Specific Relationships Among Lattice Parameters for a High Degree of Reversibility Here we answer the question of what precisely are the conditions on lattice parameters that imply the satisfaction of the proposed conditions for reversibility. It is easy to write down the conditions det U1 = 1 and λ2 = 1 in terms of the parameters in U1 , so, instead, we concentrate on the most interesting case of satisfying simultaneously all of the conditions: det U1 = 1, λ2 = 1 and the cofactor conditions. We focus only on cases where the austenite is cubic, and the symmetries have a group/subgroup relationship. In this case it is easy to see that the full set of conditions cannot be satisfied if the distortion is

170

R.D. James and Z. Zhang

such that the martensite has tetragonal or trigonal (rhombohedral), symmetry, but can be satisfied for lower symmetry martensites, as discussed later. Later, when we discuss the variants of martensite, we refer to the numbering of distortion matrices given in Sect. 9.5. 9.6.1 Cubic to Orthorhombic Transformations In the cubic to orthorhombic case there are precisely two matrices that satisfy all three conditions. These matrices are √ 1 √ ⎞ ⎞ ⎛1 ⎛1 √1 ) 1 (1 − √1 ) 0 (1 + 2) 2 ( 2 − 1) 0 (1 + 2 2 2 2 2 √ ⎟ ⎟ ⎜1 ⎜1 √ II 1 1 ⎟ ⎟ ⎜ √1 √1 UI1 = ⎜ ⎝ 2 (1 − 2 ) 2 (1 + 2 ) 0 ⎠ , U1 = ⎝ 2 ( 2 − 1) 2 (1 + 2) 0 ⎠ . √ √1 0 0 0 0 2 2 Each of these matrices satisfies the cofactor conditions simultaneously for 12 twin systems, all Type I twins for the case of UI1 and all Type II twins for the case of UII 1 (the compound twins never satisfy all three conditions). These are quite big matrices by any measure, and therefore unlikely to be realistic; they may also fall outside of the Ericksen–Pitteri neighborhood and therefore call into question the basic theory. Nevertheless, they are useful for the purpose of illustration. 9.6.2 Cubic to Monoclinic Transformations
100-Polarized In the 100 polarized cubic to monoclinic case there are two one-parameter families of matrices satisfying all three conditions, given below.  ⎞ ⎛ α2 (1 + α)(1 − α)3 0 α + α2 − α3 ⎟ ⎜ UI1 = ⎜ 1 − α2 + α3 0⎟ ⎠ , α ≤ 1 , (9.8) ⎝ α2 (1 + α)(1 − α)3 ⎛

0

(α2 + α − 1)/α2 ⎜  ⎜ 1 (α − 1)3 (α + 1) UII 1 = ⎝ α2 0

1 α2



0 (α − 1)3 (α + 1)

(α3 − α + 1)/α2 0

1 α

0

⎞

⎟ 0⎟ ⎠,

α ≥ 1 . (9.9)

1 α

Each of these works simultaneously for 12 twin systems. These are Type I twins in case (9.8) and Type II twins in case (9.9), but not all are included (i.e., there are more than 12 Type I twins in this system). Note that both families begin at the identity, corresponding to no transformation, at α = 1. The case (9.9) is particularly interesting in its proximity to real examples of distortion matrices.

9 Materials with “Unlikely” Properties

171

There are also several one-parameter families corresponding to β = 1 in the notation above. These are less interesting because they do not seem to be close to any realistic cases that could provide starting points for alloy development. On the other hand, some of these cases apply to many twin systems.
110-Polarized In the 110 polarized cubic to monoclinic case there are several one-parameter families of matrices satisfying all three conditions. We found seven such families, all passing through the identity, and all applying simultaneously to multiple twin systems. There is also a two-parameter family of matrices satisfying all three conditions, but only applying to a limited family of compound twins. While it is possible to write analytical expressions for the seven families matrices satisfying these conditions, these are somewhat complicated to write down so we do not give them here. 9.6.3 Relationships for Martensite/Martensite Transitions The conditions we propose for reversibility are applicable to martensite– martensite transitions, such as the tetragonal to trigonal transformation that occurs at a morphotropic boundary. These should imply low hysteresis by the same reasoning that we have given for the austenite/martensite transitions. If U and V denote (positive-definite, symmetric) distortion matrices corresponding to two different martensites, the condition (1) of no volume change is det U = det V, (2) the exact compatibility is that the middle eigenvalue of V−1 U2 V−1 is 1. The cofactor conditions are (9.4) and (9.5) with U1 , U2 replaced by U, V, respectively, and I replaced by either U or V. It is also possible to pass from the description (9.4) and (9.5) to more compact conditions like (9.6) in the martensite/martensite case using the methods of [14]. There are also further conditions that would make two compatible laminates of martensite additionally flexible, that would likely also enhance the reversibility of transition.

9.7 Tuning Lattice Parameters to Satisfy Two of the Proposed Conditions in the NiTiCuPd System We have recently put these ideas into practice in a special case, that of the NiTiCuPd system. It is known that a family of alloys in the system NiTiCu nearly satisfy the second condition λ2 = 1, which suggested to us that this is a good starting place. We focused on trying to determine the alloys that satisfy just the first two conditions det U1 = 1 and λ2 = 1.

172

R.D. James and Z. Zhang

Fig. 9.3. A part of the compositional triangle NiCuPd at Ti = 50%. Compositions on the line satisfy the condition λ2 = β = 1. A small subset of the measured samples is also shown. The shaded region indicates the error expected in this determination

Briefly, we measured lattice parameters of both austenite and martensite by X-ray diffraction and thereby established the mapping between composition and distortion matrices. (For a more detailed description of these experiments, see Zhang [17].) Over a large compositional region we found that det U1 = 1 within our error of measurement. However, the middle eigenvalue λ2 was more sensitive to composition.5 An example of our results is shown in Fig. 9.3. The compositional space NiTiCuPd is three-dimensional. Figure 9.3 shows a 2D slice through this surface at constant Ti composition of 50%.

9.8 Further Comparisons with Experiment After determining the mapping between composition and distortion matrices as described in Sect. 9.7, we were informed of additional data on hysteresis in the literature for alloys of particular composition. (We are grateful to Wuttig, Quandt, and Berg for these references.) 5

Also, the error of the measurement of λ2 was smaller than that of det U1 .

9 Materials with “Unlikely” Properties

173

We first describe measurements of Winzek and Quandt [18]. They measured thermal hysteresis in both free standing and stressed films, and both had similar trends, but, for definiteness, we compare with the measurements on the free standing films. Winzek and Quandt also measured hysteresis in two different ways: by using the width of a parallelogram approximation of the loop, and by using the thinnest place on the loop (some loops had a “waist”). We compare with the former method. In their measurements on NiTiCu, they found that an alloy with approximate composition Ni34 Ti50 Cu16 exhibited the lowest hysteresis among alloys they studied. We did not find any alloys in the NiTiCu system that satisfy λ2 = 1 within our experimental error. However, the alloy in the NiTiCu system that most closely satisfies λ2 = 1 is Ni30 Ti50 Cu20 , which is quite close to that of Winzek and Quandt (in fact, the closest of those they measured). We believe that if they had measured the alloy Ni30 Ti50 Cu20 , it would have exhibited the lowest hysteresis. More interestingly, Winzek and Quandt [18] also made similar measurements on NiTiPd films. They found that there was a rather sharp drop in the hysteresis at the composition Ni40 Ti50 Pd10 , among alloys they tested. This is very close to the crossing point of our surface λ2 = 1, as can be seen from Fig. 9.3; in fact, the exact crossing is at Ni39 Ti50 Pd11 . Also interesting is the data we have collected from the US Patent 5, 951, 793 shown in Table 9.3. Some of these alloys have small percentages of elements that we have not studied, and these can either be excluded from the comparison or else included (by, e.g., assigning their percent to the element for which they substitute). It is seen from Table 9.3 that there are dramatic differences in the size of the stress hysteresis. The closest alloy to our surface λ2 = 1 is the alloy shown in bold and it is indeed very close. This is the one with the lowest hysteresis.

Table 9.3. Data on stress hysteresis vs. composition for various alloys collected from US Patent 5,951,793 Ti 49.5 50 50 49.5 49.5 50 50 49.5 49.5 49 51 49

Ni 43.5 40 47 47 46.5 42.5 42 40.5 38 36 35 36

Pd 0 0 2.5 3 4 7.5 7.5 10 12 13 14 15

Other Cu 7 Cu 10 Cr 0.5

Co 0.5 V 0.5 Fe 2

Width of the hysteresis (MPa) 236 172 263 148 137 95 82 54 106 103 127 170

Af (◦ C) 25 60 55 5 5 25 15 −20 30 −30 60 −50

174

R.D. James and Z. Zhang

9.9 Summary and Outlook: A General Method for Seeking New Classes of Functional Materials We have proposed three relationships for reversibility: det U1 = 1 (and, as many compatible variants of martensite as possible), λ2 = 1, λ2 being the middle eigenvalue of U1 , and the cofactor conditions λ2 = 1,

1 tr U21 − det U21 − 2 − |a|2 > 0, 4

a · U1 cof (U21 − I)n = 0 . (9.10)

The experimental evidence rather strongly suggests that the first two of these govern the main part of the hysteresis, while the third condition is a natural extension of these concepts. The standard ideas that are usually quoted as governing hysteresis in martensitic phase transitions do not seem to us to be the most influential factors. More generally, our idea can be expressed as follows. Arrange to have a big first order, reversible (martensitic, i.e., diffusionless) phase transformation that separates two phases with different electromagnetic properties. This leads to the following advantageous situation. (1) Since the two phases have different lattice parameters, and different band structures, the possibilities for simultaneous unlikely properties are greatly improved. (2) If the phase transformation is reversible then the volume fraction of the two phases could be changed by using an electric field, a magnetic field, or a stress, depending on the shape change and electromagnetic properties of the two phases. Besides ferroelectricity – ferromagnetism, there are many potential property pairs that exhibit lattice parameter sensitivity and are candidates for the proposed strategy: solubility–insolubility of H2 , high band gap–low band gap semiconductors, insulator–conductor (electrical or thermal), opaque– transparent (at various wavelengths), high–low index of refraction, luminescent–nonluminescent. Also possible according to this strategy are new kinds of thermoelectric and thermomagnetic materials, that utilize the lattice parameter sensitivity of electromagnetic properties together with the latent heat of transformation. Acknowledgments We are grateful to K. Rabe, M. Wuttig, E. Quandt, J.M. Ball, T. Shield, J. Cui and B. Berg for stimulating discussion of these ideas. This work was supported by MURI N00014-010100761 administered by ONR. The work also benefitted from the support of NSF-NIRT DMS-0304326.

References 1. N.A. Hill: J. Phys. Chem. B 104, 6694 (2000) 2. N.A. Hill: Annu. Rev. Mater. 32, 1 (2002)

9 Materials with “Unlikely” Properties

175

3. N. Hill, K.M. Rabe: Phys. Rev. B 53, 8759 (1999) 4. R.E. Cohen: J. Phys. Chem. Solids 61, 139 (1999) 5. Y.C. Yang, X.D. Zhang, S.L. Ge, Q. Pan, L.S. Kong, H. Li, J.L. Yang, B.S. Zhang, Y.F. Ding, C.T. Ye: J. Appl. Phys. 70, 6001 (1991) 6. J. Wang, J.B. Neaton, H. Zheng, V. Nagarajan, S.B. Ogale, B. Liu, D. Viehland, V. Vaithyanathan, D.G. Schlom, U.V. Waghmare, N.A. Spaldin, K.M. Rabe, M. Wuttig, R. Ramesh: Science 299, 1719 (2003) 7. R. Tickle, R.D. James: J. Magn. Magn. Mater. 195, 627 (1999) 8. H.A. Jahn, E. Teller: Proc. R. Soc. London A161, 220 (1937) 9. R.M. Tabanli, N.K. Simha, B.T. Berg: Metall. Mater. Trans. A 32, 1866 (2001); Mater. Sci. Eng. A, 273, 644 (1999) 10. K. Otsuka, H. Sakamoto, K. Shimizu: Acta Metall. 27, 585 (1979) 11. S. Miyazaki, K. Otsuka: ISIJ Int. 29, 353 (1989) 12. R.D. James, K. Hane: Acta Mater. 48, 197 (2000) 13. K. Bhattacharya: Arch. Rational Mech. Anal. 120, 201 (1992) 14. J.M. Ball, R.D. James: Arch. Rational Mech. Anal. 100, 13 (1987) 15. J.M. Ball, R.D. James: Philos. Trans. R. Soc. London A338, 389 (1992) 16. R.V. Kohn, S. M¨ uller: Philos. Mag. A 66, 697 (1992); Commun. Pure Appl. Math. 47, 405 (1994) 17. Z. Zhang: MS thesis, University of Minnesota (2004) 18. B. Winzek, E. Quandt: Proc. Mater. Res. Soc. Symp. 604, 117 (2000)

10 Invar and Anti-Invar: Magnetovolume Effects in Fe-Based Alloys Revisited E.F. Wassermann and M. Acet

10.1 Introduction In search for a cheaper length standard than Pt–Ir, Guillaume found in 1897 “Invar,” a ferromagnetic (FM) fcc Fe65 Ni35 alloy showing small, almost temperature independent thermal expansion below the Curie temperature. After discovering “Elinvar,” an FM fcc Fe–Ni–Cr alloy with constant elastic behavior, “revolutionizing” the watch industry with newly designed balance springs, Guillaume was honored with the Nobel Price in physics in 1920. In spite of worldwide research, understanding of Invar and Elinvar remained a centennial problem until the beginning of the 1990s, when ab initio calculations gave basic insights into the relations between lattice structure, atomic volume and magnetic properties. At that time, “anti-Invar” also was recognized, the effect being the enlargement of the atomic volume of fcc Fe and fcc Fe-rich alloys at high temperatures in the paramagnetic range by FM spin fluctuations. In this chapter we briefly review the basic experimental properties of Invar and anti-Invar and discuss the properties of Fe, the element inspiring Weiss to his historical 2γ-states model. We then revisit the theoretical situation and specifically discuss the models stressing the importance of charge transfer at the Fermi energy between electronic levels with different symmetry (eg and t2g ) and nonbonding or antibonding character.

10.2 Invar In spite of intensive research for more than a century, the Invar problem remains to be only partially solved. The actual knowledge at the time being, as well as the progress in the understanding of Invar, can be traced best in some review articles [1–5], in a recent book [6] and in some key experimental [7–9] and theoretical [10–12] papers. To date, the situation can be characterized by the statement that especially concerning the behavior at finite temperatures, there is appreciable disagreement between experiment and theory. The main

178

E.F. Wassermann and M. Acet

reason is the lack of appropriate theoretical “tools” to reliably describe the finite temperature behavior of itinerant magnets from first principles. This is, however, a prerequisite to fully explain the experimentally observed Invar typical features, most significantly the temperature dependence of the thermal expansion coefficient α(T ). Invar materials show a volume increase in the magnetically ordered range below the ordering temperature, and therefore relative to a nonmagnetic lattice curve (Gr¨ uneisen curve) a reduced thermal expansion. In ferromagnetically ordered alloys below the Curie-temperature TC , this is called “FM Invar-effect,” in antiferromagnetically (AF) ordered systems below the N´eel temperature TN we get “AF Invar-effect”. Figure 10.1 shows some typical experimental examples of the thermal expansion coefficient α(T ) of fcc FM Fe65 Ni35 Invar (open circles), FM Cementite Fe3 C+2 at% Cr (full circles), and AF Ti35 Fe65 with hexagonal Laves phase C14 structure. Note that in all alloys, the FM or AF Invar effect vanishes for T > TC or T > TN and uneisen analysis. thus the reference curve (αlattice ) can be obtained by a Gr¨ The area between the Gr¨ uneisen curve and the experimental data give the maximum volume enhancement due to the Invar effect – the “spontaneous” volume magnetostriction at zero temperature ωso . The values are ωso = 1.6%

Fig. 10.1. Thermal expansion coefficient α vs. the temperature for fcc FM Invar Fe65 Ni35 , FM cementite Fe3 C + 2 at%Cr, and AF Invar Ti35 Fe65 with Laves phase C14 hexagonal structure. Dashed curve: nonmagnetic Gr¨ uneisen reference αlattice (T ). Arrows mark the magnetic ordering temperatures [6]

10 Invar and Anti-Invar

179

Table 10.1. Invar systems Structure

Material

fcc

Fe–Ni, Fe–Pt (ord.) Fe–Pt (disord.), Fe–Pd, Fe–Mn, Co–Mn, Fe–Ni–Mn, Fe–Ni–Cr, Fe–Ni–Co, Fe–Ni-Pt, Fe–Ni–Pd Cr–Fe, Cr–Mn, Fe–Cr–Mn Co–Cr Fe–B, P Fe–TE, TE: Sc, Y, Zr, Hf (Fe–TE)X TE: Mn, Co, Ni; X: Si, B, P TEFe2 TE: Ti, Zr RECo2 RE: all except Eu RE2 Fe17 RE: Y, Dy, La Dy2 (FeCo)17 , Fe3 C, Fe14 Nd2 B

bcc hexagonal amorphous

Laves phases and compounds

for FM Invar Fe65 Ni35 , ωso ≈ 1.0% for cementite and ωso ≈ 0.6% for AF Invar Fe65 Ti35 . As shown in Table 10.1, Invar behavior (and often Elinvar behavior, the temperature independent elastic behavior) is not bound to fcc Fe-based systems alone. But, because of technical importance, Fe-based alloys are most widely investigated. Applications of Invar range from large scale, e.g., Fe– Ni for ship tanks for transportation of liquified natural gas at temperatures around 110 K [4] (Fig. 10.2), to small scale in e.g., chip base plates or bi-metals.

Fig. 10.2. Large scale application of iron–nickel Invar: View into a tank of a ship for transport of liquid natural gas (capacity 130, 000 m3 , atm. pressure, T = 110 K). The tank walls are made of weldable Fe65 Ni35 Invar, specially treated to withstand repeated thermal cycling (Photo: Imphy Ugine Precision [4])

180

E.F. Wassermann and M. Acet

Fig. 10.3. Magnetic contribution cmag to the specific heat as a function of the reduced temperature for pure Ni and Fe100−x Nix alloys with different composition [13]. For details see text

Most important on the world’s Invar market, also by tonnage share, has been Fe65 Ni35 Invar used for shadow masks in television tubes. With increasing spread of flat screens, this market section of Invar is, however, dying away, none the least because the current and voltage feedthroughs of the glass TV tubes are also made of Invar (an Fe–Ni–Co alloy known as “Kovar”). Guillaume’s original Elinvar alloy (Fe54 Ni34 Cr12 ) began losing its importance a long time ago too, when Invar balance springs in watches were replaced by quartz crystals. Today, Elinvar finds applications in RHEED contacts, relays, and frequency standards – a relatively small but a high-tech market. Invar characteristic properties are by no means limited to the thermal expansion, but anomalies are found in all physical properties of the alloys mentioned, especially the magnetization M (T, H) (in the context where the problem of the “hidden excitations” is still debated; see e.g. [1–5]), the elastic constants CL,T (T ) and the bulk and shear moduli B(T ) and G(T ), the latter three revealing the characteristics of the not fully understood “softness” of the Invar materials [1–6]. Invar-typical is also the behavior of the specific heat. This is depicted for Fe–Ni alloys in Fig. 10.3 [13] showing that relative to the reference behavior of pure FM Ni, there are appreciable magnetic contributions cmag to the specific heat in FM fcc Fe100−x Nix alloys at temperatures below as well as above TC . Starting from a composition Fe50 Ni50 , this excess contribution to cmag finds its maximum at the Invar concentration x = 35. The excess magnetic specific heat is caused by thermal Invar-characteristic excitations – still debated concerning their nature as will be discussed below. These are known as moment–volume fluctuations and are also referred

10 Invar and Anti-Invar

181

to as “elasto-magnons”. Figure 10.3 reveals that there is also an appreciable magnetic contribution to the specific heat in the paramagnetic temperature range for T > TC , which is also maximum in Fe65 Ni35 Invar. This excess is very likely caused by FM spin fluctuations [14], though, again, the physical nature of the underlying excitations is not understood.

10.3 From Invar to Anti-Invar Our investigations [6–9] have shown that increasing the Fe-concentration in Fe100−x Nix alloys to above the Invar composition, x > 65 (electrons per atom, e/a < 8.75), results in a continuous transition to another magnetovolume effect called “anti-Invar” [15]. Anti-Invar is the opposite of Invar behavior and denotes the increase of the volume and the enhancement of the thermal expansion relative to a nonmagnetic reference. Anti-Invar is observed in γFe [16] and fcc Fe-rich systems in the high temperature paramagnetic range. The relative volume increase through the anti-Invar effect can be determined quantitatively by plotting Tm α(T ) as a function of T /Tm , where Tm is the melting temperature. For “normal” metals and alloys, such a plot results in an almost universal expansion curve, which we use as a Gr¨ uneisen-like reference for the thermal expansion of the magnetic alloys in question. Figure 10.4 shows such a plot for Fe100−x Nix alloys in the fcc stability range together with the data for pure γ-Fe above the A3 point (1184 K) and αlattice Tm vs. uneisen reference curve mentioned. The Fe–Ni curves terminate T /Tm , the Gr¨ at the martensite start temperatures Ms /Tm , where a transition from the high temperature austenitic fcc phase into the bcc (or bct) martensitic phase takes place on decreasing T . Very important in this context is the thermal expansion curve of Fe77 Ni13 Mn7 C3 (open circles in Fig. 10.4). This paramagnetic alloy has fcc stability in the range 0 ≤ T ≤ Tm , and thus can serve as an “anti-Invar reference” [17], i.e., it provides the course of the thermal expansion behavior α(T ) for the Fe–Ni alloys in the temperature range where the fcc phase is not stable. Note that the transition from Invar to anti-Invar behavior in Fig. 10.4 can be observed twice: with increasing Fe-concentration (at constant temperature) starting from pure fcc Fe65 Ni35 FM Invar to pure fcc anti-Invar in Fe77 Ni13 Mn7 C3 , and with increasing temperature (at constant composition), as observed in, e.g., Fe70 Ni30 through the “negative” deviation from the lattice reference at low temperatures changing to a “positive” deviation at higher temperatures. For x > 75 to pure γ-Fe, all the alloys in Fig. 10.4 are paramagnetic and show only anti-Invar behavior. Analogous transitions from Invar to Anti-Invar behavior with increasing Fe concentration as in Fig. 10.4 have also been found in Fe–Pt and Fe–Ni–Co alloys. Anti-Invar typical anomalies can be found in other physical properties of the alloys mentioned above. However, since there is no magnetic long range order, and the anomalies occur at relatively high temperatures, the number of

182

E.F. Wassermann and M. Acet

Fig. 10.4. αTm vs. Tm (melting temperature) for Fe-rich Fe–Ni alloys. The curves for 0 ≤ x ≤ 30 are measured down to their corresponding martensite start temperatures uneisen). Open circles: anti-Invar alloy Ms /Tm . Dashed : lattice reference curve (Gr¨ Fe77 Ni13 Mn7 C3 [17]

available data is limited. Attempts to find anti-Invar typical behavior in, e.g., the elastic constants (e.g., [2]) have so far not been very successful. In Fig. 10.5, we present (analogous to Fig. 10.3) our recent data on the magnetic contribution cmag to the specific heat through the anti-Invar effect for three alloys, Fe80 Ni20 , Fe74 Ni26 , and Fe77 Ni13 Mn7 C3 [13] together with some theoretical data from Monte Carlo simulations of high-moment–low-moment transitions in Fe–Ni anti-Invar [18]. There is no room to discuss the details here, but it is obvious from the figure that there are very large magnetic anomalies in these alloys (the low temperature minimum in Fe65 Ni35 Mn7 C3 results from a spin glass order), which can be described by a Schottky-type analysis that could follow from a low-spin–high-spin transition in a two level system (see also, Sects. 10.4 and 10.5). Since the thermal expansion of almost all fcc Invar and anti-Invar alloys (besides those that are Co based) is isotropic, i.e., 3(∆l/l) = ∆V /V , using the Gr¨ uneisen lattice and the Fe77 Ni13 Mn7 C3 anti-Invar references and the experimental α(T ) data, one can work out a plot demonstrating the systematic transition from Invar to anti-Invar as a function of composition in very different alloy systems. Figure 10.6 gives the result in the form of maximum relative

10 Invar and Anti-Invar

183

Fig. 10.5. (a) Anti-Invar contribution to the specific heat canti-Invar vs. temperature determined experimentally [13] for x = 20, 26 and Fe77 Ni13 Mn7 C3 anti-Invar alloys and (b) compared results of Monte Carlo simulations for a two level low-spin–highspin transition system [18]

volume enhancement (∆V /V )max vs. composition (in electrons per atom, e/a, counting s + d electrons) for a series of alloys (see symbols in Fig. 10.6). Invar behavior with a maximum volume increase of (∆V /V )max ∼ 1.6% at e/a = 8.75 diminishes with decreasing electron concentration and vanishes for e/a < 8.5. There is Invar and anti-Invar simultaneously present (also as a function of temperature – which this plot does not show directly) in a small interval 8.5 < e/a < 8.8. There is then the anti-Invar developing with decreasing e/a, with a maximum relative volume increase up to almost 6% in Fe–Mn. This relative volume increase by the anti-Invar effect occurs in addition to the relative volume increase of (∆V /V ) ∼ 7% all metals and alloys experience between 0 K and their melting temperature (Gr¨ uneisen law). The continuous transition from Invar to anti-Invar when reducing the electron concentration e/a as shown in Fig. 10.6 is directly reflected in the ab initio ground state calculations [10, 11] of the total energy as a function of atomic volume in the Fe–Ni system. Simultaneously, the composition dependence of the electronic density of states (DOS) in these fcc alloys reveals typical features

184

E.F. Wassermann and M. Acet

Fig. 10.6. Relative volume enhancement (∆V /V )max as a function of the electron concentration per atom e/a for different Invar and anti-Invar alloys (see symbols)

at energies around the Fermi energy (see also below, Sect. 10.6), which can be interpreted in terms of the long debated high-spin (HS)–low-spin (LS) model, frequently used for the microscopic explanation of both, Invar and anti-Invar. Since the volume and temperature dependences of the total energy of pure iron already exhibit these HS–LS features, an understanding of elemental Fe is a prerequisite for the discussion of different Invar and anti-Invar (and also martensite) models. Iron will therefore be discussed in Sect. 10.4.

10.4 Allotropy of Pure Fe Figure 10.7 shows the allotropy of elemental Fe in a plot of temperature vs. pressure. At atmospheric pressure, Fe solidifies in the bcc δ-phase and below 1665 K (A4 point) structurally transforms on cooling, so that for 1184 K < T < 1665 K it has the fcc structure (γ-Fe). Yet, unexpected from entropic arguments, on further cooling below A3 , Fe does not become hexagonal close packed (hcp; ε-Fe), as one would also expect from its position in the periodic table (the elements Ru and Os below Fe in column VIII are hexagonal), but the bcc structure returns (α-Fe), and remains as the stable structure down to

10 Invar and Anti-Invar

185

Fig. 10.7. Structural phase diagram of iron. For discussion see text

0 K. Thermodynamic considerations [19] show that the presence of strong FM correlations is a sufficient condition for the stability of the α-phase. Long range FM ordering, which occurs well below A3 at the Curie-temperature TC = 1040 K, is not a necessary condition. With respect to Invar and anti-Invar, the key feature of Fe is that the atomic volume of the high temperature fcc γ-phase is smaller than that of the low temperature bcc (α)-phase. Only ε-Fe occurring under high pressure is denser packed than γ-Fe, as the course of the γ–ε boundary line in Fig. 10.7 shows. If one extrapolates linearly this boundary line to zero pressure (dashed line), one obtains at ∼490 K the hypothetical γ–ε transition temperature, in the case that α-Fe were nonmagnetic (NM). The allotropy of Fe can be understood from first principles ab initio band calculations [10–12], showing that magnetic coupling in the ground state of fcc Fe could be AF or FM depending on the volume, thereby also corroborating the 2γ-states model of Weiss [20]. In Fig. 10.8, we show a recent result [21] in plots of the total energy E (bottom panel) and the magnetic moment µ (top panel) vs. the atomic volume Va for bcc, fcc, and hcp Fe. In accordance with the phase diagram of Fig. 10.7, the most stable state, i.e., lowest in total energy, is the FM bcc ground state with the minimum A). This is about equal to at Va = 77.37 a.u.3 (lattice constant a = 2.84 ˚ A). The bcc (FM) mothe experimental value Va = 76.55 a.u.3 (a = 2.83 ˚ ment changes continuously with volume and has a value of about µ = 2.2 µB at equilibrium. Note that if the bcc state were NM, the total energy curve (not shown in Fig. 10.8) would have lied about 35 mRy higher than the bcc

186

E.F. Wassermann and M. Acet

Fig. 10.8. Total energy and magnetic moment vs. volume at T = 0 K of Fe in the bcc, fcc, and hcp structures [21]

state (1 mRy = 150 K), so that NM bcc Fe is never stable. The structure with the next highest energy is hcp (NM) (dashed-dotted), situated about 5 mRy above the bcc ground state. However, the minimum lies ∼ 10 a.u.3 lower in volume, so that this state can only be reached under high pressure, in agreement with Fig. 10.7. The next stable structure in Fig. 10.8 is fcc (AF) (dashed) A). The with µ ∼ 1.5 µB at the equilibrium volume of Va = 73 a.u.3 (a = 3.51 ˚ A). This is experimental values are µ ∼ 0.7 µB and Va = 76.19 a.u.3 (a = 3.56 ˚ the state with the small moment and small volume of fcc Fe – the LS state in the 2γ-states model of Weiss [20]. Energetically, this LS-state is 7.5 mRyd (∼1125 K) above the bcc ground state, a value almost in quantitative agreement with A3 = 1180 K in Fig. 10.7. If the volume of the fcc AF LS-state is A), an fcc (FM) state is energetically enlarged, for V > 80 a.u.3 (a = 3.62 ˚ favored. In agreement with the Weiss model [20], this is the FM HS state of γ-Fe with a magnetic moment of µ ∼ 2.7 µB , a value close to µ = 2.8 µB as extrapolated from the Slater–Pauling curve. An LS state solution also exists for the fcc (FM) curve having a minimum near the equilibrium volume of the AF solution with µ ranging between 0 and 1 µB , so that µ in the FM solution changes rapidly from small values to 2.7 µB . This is known as the moment–volume instability.

10 Invar and Anti-Invar

187

Fig. 10.9. Hyperfine field Bhf as a function of the atomic volume for fcc γ-Fe anti-Invar [22]. Data from M¨ ossbauer experiments on epitaxial films grown on Cu (001) (filled square, filled circle), Cu3 Au (filled triangle), CuAu (filled diamond ), and coherent precipitates in Cu–Al (open down triangle, open square)

The step-like behavior of the moment as a function of the atomic volume has been found experimentally for fcc Fe. Results are given in Fig. 10.9 showing the hyperfine field vs. the atomic volume from low temperature M¨ ossbauer investigations on fcc Fe films epitaxially grown on substrates with different lattice constants: Cu(001) (filled square, filled circle), Cu3 Au (filled triangle), CuAu (filled diamond), and coherent precipitations of γ-Fe in Cu–Al alloys (open down triangle, open square) [22]. Since the hyperfine field is directly proportional to the magnetic moment, these data show in a convincing fashion the validity of the findings by Weiss [20]. We will demonstrate later (Sect. 10.6) that similar behavior can be detected in pressure experiments on Invar alloys.

10.5 Ground State Properties of Invar and Anti-Invar Like for pure Fe, first principles ab initio band calculations of the total energy as a function of atomic volume, magnetic moment, and composition have given insight into the ground state properties of systems like Fe–Ni or Fe– Pt [10–12]. All important is that, independent of the system, the basic features in the total energy behavior are similar. In Fig. 10.10a–d, we give the results

188

E.F. Wassermann and M. Acet

Fig. 10.10. Total energy E relative to a ground state energy E0 (full curves), and magnetic moment (in µB /atom; dashed ) as a function of the lattice constant (in atomic units a.u.) for fcc FM Fe–Ni–Co alloys with different compositions [23], showing anti-Invar properties in (a) and Invar in (b–d). For details see text

of the calculations for Fe–Ni–Co [23], a ternary system of technical importance, because it incorporates “Kovar” (composition around Fe55 Ni29 Co16 ) and “Super-Invar” (Fe64 Ni31 Co5 ), the alloy with the smallest known expansion coefficient of α = 0.2 × 10−6 K−1 at room temperature. The plots in Fig. 10.10a–d give the total energy E (relative to the ground state energy E0 ; full curves) and the magnetic moment µ (dashed curves) as a function of the lattice constant a (in a.u.) for four different compositions (Co content constant at 16.5 at%) in the transition range from anti-Invar (Fig. 10.10a; e/a = 8.625) to Invar (Fig. 10.10b–d; 8.685 ≤ e/a ≤ 8.805). Typical for both anti-Invar and Invar is a “double well potential” in the total energy curve E(a) and a “steplike” feature (moment–volume instability) in the magnetic moment. However, while in anti-Invar (Fig. 10.10a) the “anharmonicity” in E(a) is found at larger volumes than the equilibrium value V0 , and this ground state is an NM state, in Invar (Fig. 10.10c–d), the “anharmonicity” occurs at smaller volumes than V0 . The ground state is always a HS state. The change from anti-Invar to Invar with increasing concentration e/a (cf. Fig. 10.6) at zero temperature is, therefore, characterized by a reversal of the transition from NM–HS in anti-Invar to HS–NM in Invar. The energetic difference ∆E = ENM − EHS is an almost linear function of e/a [11], changing sign around e/a ∼ 8.65. The theoretical results in Fig. 10.10a–d almost agree with experiment. Only the anti-Invar behavior of the alloy in Fig. 10.10a is not very pronounced because of the occurrence of martensite [23].

10 Invar and Anti-Invar

189

The HS–NM transitions in E(a) occur at saddle points (see dotted vertical lines in Fig. 10.10, where a third state, called the LS state with an intermediate value of the magnetic moment like in fcc Fe (cf. Fig. 10.8) might be involved. However, the presence of an LS state is not a necessary condition for the occurrence of Invar or anti-Invar behavior, a two level NM–HS scheme is sufficient. Moreover, for all Invar and anti-Invar alloys studied in ab initio calculations, the energetic differences between the HS and the LS or NM states turn out to be of the order of 1 mRy. It is thus tempting to assume that by raising the temperature, “excitations” – of whatever physical nature – from the large volume high moment ground state to the energetically elevated zero (or small) moment state at smaller volumes compensate the normal Gr¨ uneisen expansion and thus cause the Invar magnetovolume anomaly; and vice versa NM–HS transitions from small to large volumes cause anti-Invar. This scenario reminds us in many ways of the long standing widely debated 2γ-states (or HS–LS states) model of Weiss [20] introduced by him to understand fcc Fe and the Invar effect at finite temperatures in Fe–Ni (the vocabulary “anti-Invar” was not yet in use at that time).

10.6 Pressure Experiments: Evidence for High Spin to Low Spin State Transitions While in anti-Invar a lattice expansion yields a step-like LS to HS state transition (see γ-Fe in Fig. 10.9), in Invar a lattice contraction with pressure should lead to the reverse, a HS to LS state transition. Abd-Elmeguid and Micklitz [24] were first to show in M¨ ossbauer investigations under high pressure at 4.2 K that indeed such a behavior could be inferred from their results on ordered and disordered Fe72 Pt28 as well as Fe68.5 Ni31.5 Invar. Unfortunately, Fe65 Ni35 cannot be investigated, because samples of this composition transform martensitically under pressure. For this reason, also recent investigations of the ac-susceptibility as a function of pressure and temperature were carried out on Fe68.1 Ni31.9 , also disordered Fe72 Pt28 and Fe70 Pt30 , and Fe66 Pd34 Invar alloys [25–27]. In Fig. 10.11a and b, we summarize all the data [24–27]. In Fig. 10.11a, the relative effective hyperfine fields Beff (p)/Beff (0) and in Fig. 10.11b, the relative Curie temperatures are plotted as a function of the pressure with compositions as indicated in the figure. Especially in Fig. 10.11a, for ordered (filled triangle) and disordered (open triangle) Fe72 Pt28 Invar, there is definitely a critical pressure necessary to induce a reduction of the magnetic moment. In Fe66 Pd34 the moment even remains almost constant up to p ∼ 5 GPa. There is also a change in slope in the Beff (p) curves of all three Invar samples, Fe68.5 Ni31.5 and Fe72 Pt28 (ordered and disordered), with the tendency towards a second plateau at high pressures. This is, to some extent, supported by the plateaus occurring in the Curie-temperature curves TC (p) in Fig. 10.11b. In disordered Fe70 Pt30 (open square) and Fe68.1 Ni31.9 (filled

190

E.F. Wassermann and M. Acet

Fig. 10.11. (a) Normalized pressure dependence of the average effective hyperfine fields at 4.2 K, Beff (p)/Beff (0) [24], and magnetization M (p)/M (0) at room temperature [25]. (b) Normalized Curie-temperatures TC (p)/TC (0) vs. pressure [24–27] for different Invar alloys as given indicated by various symbols

square), even AF components in a type of low magnetization reentrant spin glass state are found at high pressure. Though not in absolute clearness, all these data give strong support for the existence of HS and LS states and transitions within these states with pressure (at 4.2 K) in the sense of the model by Weiss [20]. Strong support for the Weiss hypothesis also stems from the pressure experiments by Odin et al. [28], who investigated Fe72 Pt28 Invar at room temperature with X-ray Magnetic Circular Dichroism (XMCD) up to 20 GPa. The result for disordered Fe–Pt is shown in Fig. 10.12 in a plot of the relative XMCD signal vs. pressure. From these data, it is clear that on reduction of the volume with increasing pressure in disordered Fe72 Pt28 Invar, there are two step-like transitions, from the initial HS state (admixed with LS states, as the authors claim) to a pure LS state and then to an NM state. This finding is in agreement with ab initio first principle calculations on Fe–Pt Invar by Podgorny [29]. However, it is to some extent in contradiction with the results of another group to what has been found on disordered Fe72 Pt28

10 Invar and Anti-Invar

191

Fig. 10.12. Normalized X-ray magnetic circular dichroism signal, XMCD(p)/ XMCD(0), vs. pressure as measured at room temperature for disordered Fe72 Pt28 Invar [28]

Invar at room temperature in the pressure dependence of the magnetization (see Fig. 10.11a (filled circle)). What are further pros and cons concerning the existence of different energetic “states” in the sense of Weiss [20] and the band structure results? 1. The calculations [10,11] predict the HS–LS (or NM) state transitions to be of first order. To some extent this is contradicted by the experiments, e.g., the temperature dependence of the specific heat or the magnetization [3–6] or the relatively smooth Invar to anti-Invar transition with composition (Fig. 10.6). 2. The LS state in Fe–Ni Invar in the Weiss model has AF order. AF order, however, is definitely not in play, neither in FM ordered Fe72 Pt28 nor in FM cementite Fe3 C Invar. The existence of an AF LS state is not a necessary condition for the occurrence of Invar. This argument is in favor of some of the band calculations revealing HS and NM states without the presence of an LS state. 3. Models explaining specific “Invars” like Fe–Ni as a weak itinerant FM or Fe–Pt as a strong itinerant FM do not provide satisfactory solutions. 4. In this context, there has been a recent view claiming that the Invar behavior in the thermal expansion and the small bulk modulus B of Fe65 Ni35 are associated with a noncollinear FM state, i.e., with canting of the moments [30, 31]. Canting, though earlier already introduced for γ-Fe and Fe72 Pt28 Invar by Uhl et al. [32], in Fe65 Ni35 Invar according to [24] led to a “wash-out” of the HS–LS features in the total energy, and to a smooth anharmonic E(V ) curve. In a recent neutron experiment with polarization

192

E.F. Wassermann and M. Acet

analysis by Cowlam and Wildes [33], it has been demonstrated, however, that Fe65 Ni35 Invar alloys are collinear ferromagnets, supporting earlier inelastic polarized neutron scattering experiments of Lynn et al. [34], where there is no indication for the presence of noncollinearity. Furthermore, the presence of a clear discontinuity observed in the pressure dependence of the bulk modulus of Fe64 Ni36 Invar in high pressure ultrasonic experiments [35] demonstrates that the HS–LS transition is more reliably explained in terms of the Weiss model rather than in terms of a smooth transition from a collinear to a noncollinear state. 5. The “softness” of Invar, i.e. the small bulk modulus as found experimentally [36] remains unexplained in the band calculations and is, thus, one of their main draw backs. In general, there is ample experimental evidence for the existence of HS, LS, and NM states in anti-Invar and Invar materials. This provides the basis for the microscopic explanation of these effects as discussed in Sect 10.7.

10.7 HS–LS Transitions in a Microscopic Picture A first idea not much recognized at its time of what could be the microscopic origin of the Invar effect was published in 1981 by Kaspar and Salahub [37]. From band structure calculations on fcc Fe–Ni clusters with different compositions, these authors predicted that in Invar, a charge transfer from antibonding (AB) majority to nonbonding (NB) minority electronic orbitals in the vicinity of the Fermi energy EF could be the reason for the transition from a large volume, large moment HS state to a low volume, low moment LS state. This result was corroborated much later in calculations for ordered (L12 ) Fe3 Ni by Entel et al. [10]. These calculations revealed that in Invar, for large atomic volume in the HS state, there is a sharp peak in the majority density of states (DOS) just below EF formed by electrons with t2g↑ -symmetry (maximum charge density in [110] direction) and AB character. This peak faces a minimum in the DOS of the minority band at EF formed by electrons with eg↓ -symmetry (maximum charge densities in [100]-direction) and NB character. In the low volume LS state, the NB eg↓ -minimum in the DOS remains at EF , while the t2g↑ -maximum shifts energetically to above EF . Although introduction of structural disorder washes out the sharp peaks in the DOS [11], calculations for disordered Fe–Ni alloys [28] demonstrated that the important features survive. The results also demonstrated how the position of these DOS peaks relative to EF varies as a function of composition, thus highlighting the differences in the DOS typical for Invar and anti-Invar when including the structural fcc to bcc transition even for martensite. For a closer discussion of what to date is called the “t2g↑  eg↓ ” scenario, we show in Fig. 10.13a and b examples of the electronic DOS [10, 11] typical

10 Invar and Anti-Invar

193

Fig. 10.13. Ab initio calculated electronic density of states for Fe62.5 Ni37.5 in the high-spin and the low-spin state [10, 11]. The arrow indicates the t2g↑  eg↓ transitions possibly taking place at finite temperatures

for a HS state (Fig. 10.13a) and LS (or NM) state (Fig. 10.13b). These DOS represent the ground state of Invar (Fig. 10.13a) and anti-Invar (Fig. 10.13b). The calculations [10,11] also revealed that in Fe3 Ni Invar, in a certain direction of the Brillouin zone at the equilibrium volume of the HS ground state, an AB t2g↑ spin band crosses with NB eg↓ bands right at EF . Concerning the charge transfer as proposed by Kaspar and Salahub [37], the band calculations [10] revealed that at the critical volume Vc , where the HS to LS state transition takes place, there is a step in the occupation of levels, i.e. a charge transfer with a depopulation of the AB t2g↑ level at the expense of the NB eg↓ level when the volume is decreased. This AB t2g↑ → NB eg↓ charge transfer causes the volume to shrink (Invar effect), because the internal electronic pressure favoring a large volume is reduced. The opposite, i.e., an NB eg↓ → AB t2g↑ charge transfer, causes the pressure to rise and thus the volume to increase (anti-Invar effect).

194

E.F. Wassermann and M. Acet

Support for the validity of this t2g↑  eg↓ scenario comes from various sources: •

First, within this picture, it is not necessary to differentiate between weak and strong itinerant FM Invar and to assume different and system specific physical reasons for the occurrence of Invar and anti-Invar behavior. The experimental systematics (cf. Fig. 10.6) demonstrating the continuous transition from Invar to anti-Invar with a variation of the electron concentration (e/a) is naturally explained by accompanying changes in the DOS features for all alloy systems. • Second, the pressure experiments prove the existence of HS, LS and NM states, theoretically revealed in the ab initio ground state calculations. The experiments show that these states are even present at finite temperatures. • Third, the t2g↑  eg↓ scenario is corroborated by our [39] high temperature paramagnetic neutron scattering investigations on Fe100−x Nix alloys within the fcc stability range. Figure 10.14 shows the results in a plot of the magnetic moment µnorm normalized to the value of fcc Fe given in [40] vs. the temperature T for compositions between x = 35 (Invar) through the anti-Invar range (10 ≤ x ≤ 25) and to pure γ-Fe, for which we use earlier data from the literature [40]. Note that with increasing Fe concentration the slope of the curves in Fig. 10.14 changes its sign. While the

Fig. 10.14. Temperature dependence of the normalized magnetic moment as observed in Fe–Ni alloys in the paramagnetic range [39]

10 Invar and Anti-Invar

195

decrease of µnorm with increasing T occurs in all FM materials, the increase of the average moment with temperature as observed for Fe90 Ni10 and Fe77 Ni13 Mn7 C3 anti-Invar is logically explained by an eg↓ → t2g↑ charge transfer. This enhances with rising temperature the DOS of the local band within the ferromagnetic short range correlations, thus enhancing the amplitude of the longitudinal FM spin fluctuations. This simultaneously supports the LS–HS transition picture. • Fourth, we mention that recently, the zero thermal expansion of YbGaGe was found to be of electronic origin, a temperature-induced valence transition in the Yb atoms [41].

10.8 Questions and Outlook The t2g↑  eg↓ scenario as described has certain weaknesses and drawbacks raising doubts and further questions. On the one hand, recent polarized neutron experiments trying to find evidence for the eg↓  t2g↑ charge transfer to take place with temperature in Fe65 Ni35 [42] and ordered Fe72 Pt28 [43, 44] Invar have not been successful. The changes in the eg electron concentration in the temperature range from 100 to 600 K were found to be on the order of a percent in both systems, an amount probably not sufficient to explain the Invar effect on the basis of a charge transfer. On the other hand, these experiments for both alloys showed the existence of some additional excitation, a “forbidden” mode with the same dispersion as the TA1 [110] phonon but with mixed phonon–magnon character. This could be what can be called the “elasto-magnon,” the unknown HS to LS state excitation. Undoubtedly, it has been always one of the main drawbacks of the t2g↑  eg↓ scenario that the explanation of the magnetovolume effects Invar and anti-Invar did not include phonons. Entel et al. [10] speculated about the electron–phonon coupling through the formation of a “pseudo-gap” pinning the eg↓ -DOS minimum at the Fermi energy and holding it there when the temperature is increased. This has, however, not yet been confirmed experimentally. Instead, theory revisited the disordered local moment (DLM) picture, which was earlier used to explain finite temperature properties of Fe–Ni alloys [45] and Invar [46]. Within this model, an originally binary alloy system such as Fe100−x Nix is turned into a pseudo ternary alloy by differentiating between Fe-spin up and Fe-spin down atoms, (Fe1−c ↑Fec ↓)100−x Nix , in a fashion that at T = 0 K, all spins are up (c = 0), while at T = TC , 50% of the Fe-moments are pointing up and the other 50% point down (c = 0.5), which means that the magnetization has vanished [47]. Within this picture, a recent study [48] claims that the Invar effect and the “hidden” excitations are now understood, at least for disordered Fe–Pt Invar. For Fe–Ni alloys, where both types of atoms carry a moment, the LSD formalism is less straightforward. This led to the development of a magnetochemical model, in which the nonequilibrium local chemical order on an fcc Fe–Ni lattice gives rise to the Invar anomalies [49].

196

E.F. Wassermann and M. Acet

We think that as long as these models can neither explain the FM nature of the short range correlations found in the paramagnetic range in Fe–Ni Invar and anti-Invar [39] nor reveal the physical nature of the “elasto-magnons” [42–44] in a general picture for all Invar and anti-Invar type materials, more research from theoretical and experimental side is necessary to fully solve the “Invar problem,” which is, concerning the time scale now, older than a century.

References 1. The physics and applications of invar alloys. In: Honda Memorial Series on Materials Science No. 3 (Marunzen, Tokyo 1978) p. 1 2. E.F. Wassermann: Invar, moment–volume-instabilities in transition metals and alloys. In: Ferromagnetic Materials, ed. by K.H.J. Buschow, E.P. Wohlfahrt, Vol. 5 (Elsevier, Amsterdam 1990) pp. 237–322 3. M. Shiga: Invar alloys. In: Materials Science and Technology, ed. by K.H.J. Buschow, E.P. Wohlfahrt, Vol. 3b (VCH, Weinheim 1994) pp. 159–210 4. G. B´eranger et al. (Eds.): The Iron–Nickel Alloys (Lavoisier, Cachan, France, 1996) 5. J. Wittenauer (Ed.): The Invar Effect: A Centennial Symposium (TMS Society, Warrendale, PA, USA, 1997) 6. W. Pepperhoff, M. Acet: Constitution and Magnetism of Iron and its Alloys (Springer, Berlin Heidelberg New York 2001) 7. E.F. Wassermann, K. Usadel, D. Wagner (Eds.): Proceedings of the international symposium on magnetoelasticity ISOMES I (Duisburg, Germany, 1989) Physica B 161, (1989) 8. E.F. Wassermann, M. Acet, P. Entel, W. Pepperhoff: J. Magn. Soc. Jpn. 23, 385 (1999) 9. E.F. Wassermann (Ed.): Proceedings of the international symposium on magnetoelasticity ISOMES II (Duisburg, Germany, 1998), Philos. Mag. B 80, (2000) 10. P. Entel, E. Hoffmann, P. Mohn, K. Schwarz, V.L. Moruzzi: Phys. Rev. B 47, 8706 (1993) 11. M. Schroeter, H. Ebert, H. Akai, P. Entel, E. Hoffmann, G.G. Reddy: Phys. Rev. B 52, 188 (1995) 12. H. Herper, E. Hoffmann, P. Entel: Phys. Rev. B 60, 3839 (1999) 13. J. van Lier: Thesis, Universitaet Duisburg, Fak. IV (2002) 14. B. Rellinghaus, J. Kaestner, T. Schneider, E.F. Wassermann, P. Mohn: Phys. Rev. B 51, 2983 (1995) 15. T. Schneider, M. Acet, B. Rellinghaus, E.F. Wassermann, W. Pepperhoff: Phys. Rev. B 51, 8917 (1995) 16. M. Acet, H. Zaehres, E.F. Wassermann, W. Pepperhoff: Phys. Rev. B 49, 6012 (1994) 17. M. Acet, T. Roessel, E.F. Wassermann, K.H. Andersen, J. Kulda, A.P. Murani, A. Wildes: J. Phys. Soc. Jpn. 69 (Suppl. A), 108 (2000) 18. M.E. Gruner, P. Entel: Comp. Mater. Sci. 10, 230 (1998) 19. M. Acet, H. Herper, P. Entel, E.F. Wassermann: J. Phys. IV 11, Pr8 (2001) 20. R.J. Weiss: Proc. Phys. Soc. London 82, 281 (1963)

10 Invar and Anti-Invar

197

21. P. Entel, H. Herper, M. Schroeter, E. Hoffmann, K. Kadau, R. Meyer, in: Displacive Phase Transformations, ed. by K. Inoue, K. Mukharjee, K. Ohtsuka, H. Chen (The Minerals, Metals & Materials Society, 1998) 22. T. Shinjo, W. Keune: J. Mag. Magn. Mater. 200, 598 (2000) 23. B. Gehrmann: Thesis, University of Duisburg, Fak. IV (1999) 24. M.M. Abd-Elmeguid, H. Micklitz: Physica B 161, 17 (1989); Phys. Rev. B 40, 7395 (1989) 25. M. Matsushita, T. Nishimura, S. Endo, M. Ishizuka, F. Ono: J. Phys.: Condens. Matter 14, 10 753 (2002) 26. M. Matsushita, S. Endo, K. Miura, F. Ono: J. Mag. Magn. Mater. 260, 371 (2003) 27. M. Matsushita, S. Endo, K. Miura, F. Ono: J. Mag. Magn. Mater. 265, 352 (2003) 28. S. Odin, F. Baudelet, J.P. Itie, A. Polian, S. Pizzini, A. Fontaine, C. Giorgetti, E. Dartyge, J.P. Kappler: J. Appl. Phys. 83, 7291 (1998) 29. M. Podgorny: Physica B 161, 105 (1989) 30. M. van Schilfgaarde, I.A. Abrikosov, B. Johansson: Nature 400, 46 (1999) 31. P. Mohn: Nature 400, 18 (1999) 32. M. Uhl, L.M. Sandratskii, J. Kuebler: Phys. Rev. B 50, 291 (1994) 33. N. Cowlam, A.R. Wildes: J. Phys.: Condens. Matter 15, 521 (2003) 34. J.W. Lynn, N. Rosov, M. Acet, H. Bach: J. Appl. Phys. 75, 60 069 (1994) 35. F. Decremps, L. Nataf: Phys. Rev. Lett. 92, 157 204-1 (2004) 36. L. Ma˜ nosa, G.A. Saunders, H. Rahdi, U. Kawald, J. Pelzl, H. Bach: J. Phys.: Condens. Matter 3, 2273 (1991) 37. J. Kaspar, D.R. Salahub: Phys. Rev. Lett. 47, 54 (1981) 38. P. Entel, H.C. Herper, E. Hoffmann, G. Nepecks, E.F. Wassermann, M. Acet, V. Crisan, H. Akai: Phil. Mag. B 80, 141 (2000) 39. M. Acet, E.F. Wassermann, K. Andersen, A. Murani, O. Sch¨ arpff: Europhys. Lett. 40, 93 (1997) 40. P.J. Brown, I. Jassim, K.U. Neumann, K.R.A. Ziebeck: Physica B 16, 9 (1989) 41. J.R. Salvador, F. Guo, T. Hogan, M.G. Kanatzidis: Nature 425, 702 (2003) 42. P.J. Brown, K.U. Neumann, K.R.A. Ziebeck: J. Phys.: Condens. Matter 13, 1563 (2001) 43. P.J. Brown, T. Kanomata, M. Matsumoto, K.U. Neumann, K.R.A. Ziebeck: J. Mag. Magn. Mater. 242–245, 781 (2002) 44. P.J. Brown, T. Chaterji, J. Kaestner, K.R.A. Ziebeck: J. Phys.: Condens. Matter 16, 5349 (2004) 45. H. Akai, P.H. Dederichs: Phys. Rev. B 47, 8739 (1993) 46. D.D. Johnson, W.A. Shelton, in: The Invar Effect: A Centennial Symposium, ed. by J. Wittenauer (TMS Society, Warrendale, PA, USA, 1997) p. 63 47. S. Khmelevskyi, P. Mohn: Phys. Rev. B 68, 214 412 (2003) 48. S. Khmelevskyi, I. Turek, P. Mohn: Phys. Rev. Lett. 91, 037 201-1 (2003) 49. V. Crisan, P. Entel, H. Ebert, H. Akai, D.D. Johnson, J.B. Staunton: Phys. Rev. B 66, 014 416 (2002)

11 Magnetocaloric Effect Associated with Magnetostructural Transitions V.K. Pecharsky and K.A. Gschneidner, Jr.

11.1 Introduction The discovery of the giant magnetocaloric effect in Gd5 (Si4−x Gex ) in 1997 stimulated a considerable growth of research in order to both find new materials where the effect is just as potent and to understand the role magnetostructural transitions play in its enhancement. Results obtained to date indicate that the coupling between magnetic and crystal lattices is critical in achieving strong magnetothermal responses in weak magnetic fields and that advanced magnetocaloric materials should exist in solids where structural changes are linked to ferromagnetic ordering and disordering, and therefore, can be triggered by a magnetic field.

11.2 Magnetic Cooling or Why Having a Strong Magnetocaloric Effect in a Weak Magnetic Field Makes a Difference? Magnetic cooling has a potential to reduce global energy consumption and eliminate or minimize the need for ozone depleting and greenhouse chemicals. It may soon become an alternative to vapor-compression technology. Instead of a working fluid undergoing a liquid–vapor transition in a conventional cooling system, a magnetic refrigerator (see Fig. 11.1) employs a solid, which heats up when magnetized and cools down when demagnetized. Originally measured in iron and reported by Warburg in 1881 [1], these magnetic fieldinduced temperature variations are known today as the magnetocaloric effect (MCE). Intrinsic to every magnetic solid, the MCE has exceptional fundamental importance because it bridges length, energy and time scales spanning over many orders of magnitude: from quantum mechanics to micromagnetics, from statistical to macroscopic thermodynamics and from spin dynamics to heat transfer [2–11].

200

V.K. Pecharsky and K.A. Gschneidner, Jr.

Fig. 11.1. Laboratory prototype magnetic refrigerator designed and constructed by the Astronautics Corporation of America in 2001 [18]. A ∼1.5 T magnetic field around the magnetocaloric wheel filled with Gd spheres is produced by a permanent magnet. The refrigerator operates at ambient conditions with a maximum temperature span of ∼20◦ C and maximum cooling power of 95 W. The photograph is courtesy of the Astronautics Corporation of America, 4115 N. Teutonia Avenue, Milwaukee, WI 53209

In addition to its role in basic science, the magnetocaloric effect has a substantial practical potential since its extent is one of the most critical parameters defining the performance of a magnetic refrigerator – the stronger the MCE, the higher the efficiency of the device all other things being equal [11–19]. As is often the case with emerging technologies, numerous applications of magnetic refrigeration may stem from a reliable foundation formed by advanced materials. For example, the availability of low-cost highperformance solids exhibiting enhanced MCE between ∼250 and ∼350 K is an important requirement in order to facilitate commercialization of magnetic refrigeration for a variety of consumer uses – from home appliances to climate control in motor vehicles. Also, when suitable magnetocaloric compounds supporting continuous magnetic cooling from ∼300 to ∼20 K are developed, the energy penalty incurred during hydrogen liquefaction using the conventional gas-compression approach may no longer be a limiting factor preventing the widespread use of liquid hydrogen fuel in transportation. The list of materials capable of satisfying future magnetocaloric requirements is, however, limited and if a solid is needed with a certain MCE to meet a set of specific constraints that lie beyond our current state of knowledge, one often relies on an accidental discovery rather than on design guided by a rational theory.

11 Magnetocaloric Effect Associated with Magnetostructural Transitions

201

Fig. 11.2. The total entropy in zero and 7.5 T magnetic fields (a) and the magnetocaloric effect of the elemental Gd in the vicinity of its Curie temperature, TC , for a magnetic field change from 0 to 7.5 T (b). The data are for a high purity single crystalline Gd with the magnetic field vector parallel to the [0001] crystallographic direction of the specimen

As illustrated in Fig. 11.2, the magnetocaloric effect can be expressed as either or both the isothermal entropy change, ∆SM (an extensive quantity) or the adiabatic temperature change, ∆Tad (an intensive parameter). For a given material at constant pressure, the two quantities are functions of absolute temperature (T ) and magnetic field change (∆B = Bf − Bi ), where Bf and Bi are the final and initial magnetic fields, respectively. The MCE can be easily computed (see [16, 20] and Fig. 11.2a) provided the behavior of the total entropy (S) of a compound is known as a function of temperature in both the initial and final magnetic fields ∆SM (T, ∆B)∆B=Bf −Bi = S(T, B)B=Bf − S(T, B)B=Bi ,

(11.1)

∆Tad (T, ∆B)∆B=Bf −Bi = T (S, B)B=Bf − T (S, B)B=Bi .

(11.2)

Equation (11.2) is also straightforwardly employed in direct measurements of ∆Tad [21–23]. At equilibrium, both ∆SM and ∆Tad are correlated with magnetization (M ), magnetic flux density (B), heat capacity at constant pressure (C), and absolute temperature by one of the following fundamental

202

V.K. Pecharsky and K.A. Gschneidner, Jr.

relationships [24]: 

Bf

∆SM (T, ∆B)∆B=Bf −Bi = Bi

 ∆Tad (T, ∆B)∆B=Bf −Bi = −

Bf

Bi





∂M (T, B) ∂T

 dB ,

(11.3)

B

T ∂M (T, B) × C(T, B) ∂T

 dB .

(11.4)

B

As immediately follows from (11.1) to (11.4), materials whose total entropy is sharply influenced by a magnetic field and where magnetization changes rapidly with temperature, are expected to exhibit enhanced MCE. The latter peaks when |(∂M (T, B)/∂T )B | is the greatest, i.e., around the Curie temperature (TC ) in a conventional ferromagnet or near the absolute zero temperature in a paramagnet. The MCE usually decreases both below and above the TC , as clearly seen in Fig. 11.2b. To amplify the MCE, conventional wisdom calls for increasing the magnetic field change in addition to maximizing both the derivative, |(∂M (T, B)/∂T )B |, and the region of magnetic fields and temperatures where the magnetization remains highly sensitive to temperature. Considering the current state-of-the-art in permanent magnet materials, however, it is quite unlikely that magnetic fields in excess of about 2 T will become common in a foreseeable future. That is why maximizing the MCE by manipulating one or several most critical parameters of a compound appears to be the most realistic option in order to reach the strongest possible MCE’s in readily available magnetic fields. Although it remains a formidable challenge for basic science, a better understanding, and therefore, the ability to control a variety of chemical, structural, and physical degrees of freedom that define the properties of complex solids may improve the existing materials and should eventually result in novel compounds exhibiting large magnetocaloric effects.

11.3 Gd5 (Si4−xGex) System and the Giant Magnetocaloric Effect The use of the term “giant” referring to the “magnetocaloric effect” can be traced back to 1973 [25], yet the phenomenon was not given much consideration before 1997 when Pecharsky and Gschneidner [26–28] observed the unusually potent MCE in Gd5 (Si2 Ge2 ) and other Gd5 (Si4−x Gex ) alloys with x ≥ 2, Fig. 11.3. Between 1973 and 1997, there were a few accounts of a strong MCE, the most notable examples were FeRh [31, 32], (Hf0.83 Ta0.17 )Fe2±x [33], and (La1−x Cax )MnO3 [34]. The larger than average MCE’s reported before 1997 were primarily attributed to magnetic field induced transformations involving two magnetic states: low magnetic field antiferromagnetic and high magnetic field ferromagnetic (e.g., FeRh), or to large phase volume changes

11 Magnetocaloric Effect Associated with Magnetostructural Transitions

203

Fig. 11.3. The giant MCE of Gd5 (Si2 Ge2 ) in the as-prepared and annealed (1570 K, 1 h) states: (a) −∆SM calculated from magnetization data using Eq. (11.3) [29]; (b) −∆Tad calculated from heat capacity data using Eq. (11.2) (lines) and measured directly during increasing (solid symbols) and decreasing (open symbols) magnetic fields [30]

(La1−x Cax MnO3 ). In addition to a tendency for irreversibility of the effect in FeRh, this and other systems available before 1997 were not flexible enough to enable easy tuning of the temperature range where the enhanced MCE was observed. Consequently, there was much uncertainty whether or not these materials could be used in a practical magnetic refrigerator in future. The R5 T4 compounds (R = rare earth metal and T = Si and/or Ge) have been known since 1966 [35–37]. For the next 30 years, they received little attention despite a striking observation made by Holtzberg et al. [37] that instead of the expected reduction of the Curie temperature of the elemental Gd (TC = 293 K), the ferromagnetic ordering temperature of the intermetallic compound Gd5 Si4 , in which nearly half of magnetic Gd atoms have been replaced with nonmagnetic Si atoms, was increased by 43 K to TC = 336 K. Among several other features noted by Holtzberg et al. [37], the following two were quite unusual: (i) the closely structurally related germanide, Gd5 Ge4 , orders antiferromagnetically at TN = 15 K, while its paramagnetic Curie temperature (θp = 94 K) was large and positive and (ii) upon substitution of Si for its electronic twin Ge in the pseudobinary Gd5 (Si4−x Gex ) system, no continuous solubility was observed, instead, a new phase emerges for x between 2 and 2.8. Studies performed 30 years later [38–40] indicated that the room temperature crystal structures of Gd5 Si4 and Gd5 Ge4 are different, yet they are intimately related to one another. The major difference is in the interactions

204

V.K. Pecharsky and K.A. Gschneidner, Jr.

Fig. 11.4. Crystallography of the Gd5 T4 system: the Gd5 Si4 -type where all slabs are linked via T–T bonds (the bonding distances shown as T–T dumbbells are ∼2.5 ˚ A); the Gd5 (Si2 Ge2 )-type with the slabs connected into pairs, i.e., with half of the interslab T–T bonds broken; and the Sm5 Ge4 -type where no interslab T–T bonds exist. The interslab T–T distances increase from the bonding ∼2.5 ˚ A to nonbonding ∼3.5 ˚ A as the slabs slide by ∼0.2 to ∼0.4 ˚ A in the directions indicated by short open arrows. To date, the Gd5 (Si2 Ge2 )-type ↔ Sm5 Ge4 -type transitions have been induced only by varying composition; the other two transformations can be triggered by any one or a combination of the listed thermodynamic variables

between the well-defined subnanometer-thick layers (or two-dimensional slabs) of tightly bound atoms as illustrated in Fig. 11.4. From x = 0 to x ∼ = 2, Gd5 (Si4−x Gex ) alloys maintain orthorhombic Gd5 Si4 -type structure at room temperature [38]. When Ge concentration exceeds that of Si, half of the interslab T–T bonds break and monoclinic Gd5 (Si2 Ge2 )-type structure becomes stable at room temperature for x from ∼1.9 to ∼2.5 [38, 41]. Further increasing the Ge concentration eliminates all interslab T–T bonds and a different orthorhombic structure (Sm5 Ge4 -type) emerges at room temperature when 2.8 ≤ x ≤ 4 [38, 42]. Soon after the original reports on the giant MCE of Gd5 (Si2 Ge2 ), Morellon et al. [43, 44] showed that a transition from the Gd5 (Si2 Ge2 )-type structure to the Gd5 Si4 -type structure coincides with the temperature-induced ferromagnetic ordering of the Gd5 (Si4−x Gex ) alloy with x = 2.2. A number of studies that followed [39, 40, 45–53] confirmed that ferromagnetically ordered Gd5 (Si4−x Gex ) alloys always adopt the Gd5 Si4 -type structure, while both the magnetically disordered and antiferromagnetically ordered compounds may exist in any of the three structure types illustrated in Fig. 11.4.

11 Magnetocaloric Effect Associated with Magnetostructural Transitions

205

Although diffraction studies were only performed as functions of temperature in zero magnetic fields [39, 43, 45, 47, 52], thermal expansion and magnetostriction data measured as functions of temperature, magnetic field, and/or pressure [43, 49, 50, 53], were taken as indications that the same structural transition can be triggered by either or both magnetic field and pressure in addition to temperature (see Fig. 11.4). Hence, the existence of the giant magnetocaloric effect in Gd5 (Si4−x Gex ) alloys has been implicitly related to the coupling of magnetic and crystallographic sublattices in these materials. It is worth noting that the coupled magnetostructural transitions in the Gd5 T4 system are thermodynamically first-order transformations, and therefore, the behavior of the MCE as a function of temperature is expected to be different from that illustrated in Fig. 11.2. We refer the reader to the analysis presented by Pecharsky et al. [54], although by comparing Fig. 11.2b with Fig. 11.3 it is easy to see that regardless of the differences in the magnetic field values and apart from sharp spikes1 observed in the behavior of ∆SM , the maximum MCE in Gd5 (Si2 Ge2 ) extends over a broader temperature range but then falls off more quickly when compared to the same in Gd. As ∆B increases, the temperature range over which the MCE remains large and nearly constant in first-order phase transition materials also increases, while the magnitude of the maximum magnetocaloric effect is affected to a much lesser degree [54]. As we noted above, the discovery of the giant magnetocaloric effect in Gd5 (Si4−x Gex ) alloys catalyzed further research. As a result, several different families of materials have been recently reported to also exhibit the giant MCE. These include: a host of colossal magnetoresistive manganites (e.g., [17, 55]), MnAs and MnAs1−x Sbx [56],2 MnFe(P1−x Asx ) [59], and La(Fe13−x Six ) [60] compounds.2 Changes of the magnetic structures in these materials are coupled with modifications of the crystal lattices. In manganites, crystallographic symmetry often varies similarly to that observed during the Gd5 (Si2 Ge2 )-type to the Gd5 Si4 -type transition, while in MnAs, MnFe(P1−x Asx ), and La(Fe13−x Six ), only unit cell volumes are altered. The latter, though comparable to the Sm5 Ge4 -type ↔ Gd5 Si4 -type transition observed in Ge-rich Gd5 T4 materials, is notably different because a martensitic-like structural transformation accompanies the unit cell volume changes without affecting symmetry in the Gd5 (Si4−x Gex ) system (see Fig. 11.4 and [38, 40]). Reminiscent of R5 T4 systems, structural changes in other materials exhibiting the giant MCE are normally assumed to be

1

Spikes observed in Fig. 11.3a are experimental artifacts likely originating from a combination of data collection and processing errors, the greatest being numerical integration of magnetization data measured with finite temperature and magnetic field steps.

2

We note that both the MCE and temperature-induced polymorphism in MnAs were reported as far back as 1958 [57]. Similarly to R5 T4 materials, the La(Fe13−x Six ) phases are known since 1968 [58], i.e., long before their magnetocaloric properties were measured.

206

V.K. Pecharsky and K.A. Gschneidner, Jr.

triggered by a magnetic field as deduced from the combination of temperaturedependent zero magnetic field diffraction data and temperature- and magnetic field-dependent measurements of bulk physical properties, such as thermal expansion and/or magnetostriction. The Gd5 (Si4−x Gex ) system remains the most studied to date, both with respect to its magnetocaloric effect and understanding the interplay between structure, magnetic, and thermodynamic properties of these materials. Therefore, below we will concentrate on several different Gd5 T4 stoichiometries in order to illustrate the role structural transitions play in enhancing the MCE in low magnetic fields. We will examine how the magnetic field alters the crystal structure of these materials, which by itself is an intriguing phenomenon that is not nearly as common as temperature or pressure induced polymorphism, and also evaluate the relative contributions from both crystallographic and magnetic phase changes into the appearance of the giant magnetocaloric effect in the Gd5 (Si4−x Gex ) system.

11.4 Altering Crystal Structures with a Magnetic Field One of the most interesting compounds in the Gd5 (Si4−x Gex ) system is the germanide, Gd5 Ge4 . According to a variety of experimental data [38, 53, 61– 68], it orders antiferromagnetically at TN ∼ = 130 K and in low magnetic fields, the antiferromagnetic ground state is preserved down to 1.8 K. When Gd5 Ge4 is exposed to a magnetic field exceeding ∼1 T, however, a ferromagnetic state can be induced irreversibly below 10 K, partly reversibly between 10 and 20 K and fully reversibly above 20 K. Critical magnetic field, BC , needed to induce the ferromagnetic state behaves differently in different temperature regions. Below 10 K it increases as temperature decreases; above 20 K, BC rises with the increasing temperature, and the critical magnetic field remains constant from ∼10 to ∼20 K [65]. As shown recently by Pecharsky et al. [63], magnetism of Gd5 Ge4 is closely related to its crystallography. As long as the compound remains antiferromagnetic, it maintains the Sm5 Ge4 -type structure and only lattice parameters of the material (see Fig. 11.5) are anisotropically reduced upon cooling from 300 to 5 K in a zero magnetic field. Between ∼90 and 300 K, all unit cell dimensions vary nearly linearly with the following coefficients of linear thermal expansion: 6.90 × 10−6 , 1.18 × 10−5 , and 1.03 × 10−5 K−1 along the a-, b-, and c-axis, respectively [69]. These values are in agreement with αl = (dl/dT )/l = 1.03 × 10−5 K−1 in the same temperature range as determined dilatometrically using polycrystalline Gd5 Ge4 [53, 64, 70]. We note that along the a-axis, i.e., parallel to the direction along which the slabs slide during a structural change (see Fig. 11.4), thermal expansion is considerably smaller than that along either the b- or c-axis. The volumetric thermal expansion between ∼90 and 300 K varies with αV = (dV /dT )/V = 2.91×10−5 K−1 . Thermal expansion along the a-axis becomes nearly negligible below ∼100 K,

11 Magnetocaloric Effect Associated with Magnetostructural Transitions

207

Fig. 11.5. Fragments of X-ray powder diffraction data of Gd5 Ge4 collected using Mo Kα radiation during cooling from 300 to 5 K in a zero magnetic field (left) and temperature dependencies of the unit cell dimensions determined from X-ray powder diffraction data (right). The lines drawn through the data points on the right are guides to the eye

while it approaches zero only below ∼30 K along the other two principal crystallographic directions. No obvious anomalies of thermal expansion have been detected around the N´eel temperature, TN = ∼130 K. A magnetic field has a profound effect on the crystal structure of Gd5 Ge4 , as evidenced by the X-ray powder diffraction data depicted in Figs. 11.6 and 11.7. Isothermal exposure to an increasing magnetic field at 6.1 K results in a structural transition that begins around 1.5 T and is nearly complete at 2.5 T. The magnetic field-induced structural transition in Gd5 Ge4 is irreversible at this temperature, as can be concluded from the powder patterns that barely change when the magnetic field is removed (Fig. 11.6). Thus, after magnetizing and subsequent demagnetizing, Gd5 Ge4 retains its high magnetic field crystal structure. When the magnetic field is cycled at this temperature again, the crystal structure of Gd5 Ge4 remains unaffected, which is consistent with a variety of bulk physical property measurements, including magnetization [62, 64], electrical resistivity [61], thermal expansion [64], and magnetostriction [53]. Dependence of the Gd5 Ge4 crystal structure on magnetic field is different from the above behavior when the magnetic field is cycled isothermally at 25 K. Just as at 6.1 K, the low magnetic field crystal structure transforms into the high magnetic field polymorph between 1.5 and 2.5 T (Fig. 11.7). But when the magnetic field is reduced at 25 K, the structural transformation becomes reversible and the zero magnetic field

208

V.K. Pecharsky and K.A. Gschneidner, Jr.

Fig. 11.6. Fragments of the X-ray powder diffraction patterns of Gd5 Ge4 collected at 6.1 K with the magnetic field varying from 0 to 3.5 T and then back to 0 with a 0.5 T step

Fig. 11.7. Fragments of the X-ray powder diffraction patterns of Gd5 Ge4 collected at 25 K with the magnetic field varying from 0 to 3.5 T and then back to 0 with a 0.5 T step

11 Magnetocaloric Effect Associated with Magnetostructural Transitions

209

diffraction patterns are restored between 1.5 and 0.5 T (Fig. 11.7), displaying a ∼1 T hysteresis. As follows from the results of the structure determination using powder diffraction data [63], the low magnetic field crystal structures of Gd5 Ge4 observed at 6.1 and 25 K are identical to one another within the accuracy of the technique, and so are the high magnetic field polymorphs. The low field Gd5 Ge4 adopts the Sm5 Ge4 -type structure, which is transformed into the Gd5 Si4 -type (see Fig. 11.4) by the magnetic field. The two crystal structures are related to one another via shear displacements of the neighboring layers by more than 0.2 ˚ A each. These large shifts in opposite directions alter numerous interatomic distances (Fig. 11.8), and therefore, interactions among atoms from adjacent layers. The most prominent magnetic field induced change occurs in the interlayer Ge–Ge bonds. In the low field-Gd5 Ge4 – the corresponding interatomic A – the bonds are weak but in the high fielddistances are δGe–Ge = 3.62(1) ˚ A, they become much stronger due to a 1 ˚ A Gd5 Ge4 , where δGe–Ge = 2.62(1) ˚ contraction. Despite substantial differences between the two structures, the layers themselves undergo little change, as evidenced by small variation of the interatomic distances inside each layer (see [63, 67], and Fig. 11.8). The magnetic field, therefore, triggers a martensitic-like structural transition, which results in the breaking (in low magnetic fields) and the reforming (in high magnetic fields) of the covalent-like interlayer Ge–Ge bonds. This structural change is practically identical to that induced by the varying

Fig. 11.8. Fragments of the low field (antiferromagnetic) and high field (ferromagA (after [63]). netic) crystal structures of Gd5 Ge4 with selected distances labeled in ˚ The intraslab distances vary by a maximum of ∼5%, while the interslab distances change up to ∼28% during the transformation. The open arrows indicate the directions and the magnitudes of the shifts of the slabs during the respective transitions (from low to high field and vice versa)

210

V.K. Pecharsky and K.A. Gschneidner, Jr.

Fig. 11.9. Left – normalized magnetization (lines) and concentration of the high field Gd5 Ge4 structure (symbols) [63]; right – unit cell dimensions of the majority phase at 6.1 K [69]. Short straight arrows indicate the direction of the magnetic field change

stoichiometry in the Gd5 (Si4−x Gex ) system, i.e., when x decreases from x = 4 to x = 0 and the large germanium atoms are substituted with the smaller Si atoms. Hence, the application of the magnetic field in the case of Gd5 Ge4 resembles chemical pressure effect when the phase volume, and consequently, interatomic distances are systematically reduced by chemical substitutions. The irreversibility and reversibility of the magnetic field induced structural transition in Gd5 Ge4 closely follows the irreversible and reversible behavior of the magnetization and the underlying magnetic structure changes of the compound. As illustrated in Fig. 11.9 (left), the fraction of the high field-Gd5 Ge4 structure formed at 6.1 K by an increasing magnetic field follows the initial magnetization path. Upon removal of the magnetic field isothermally, Gd5 Ge4 remains ferromagnetic at 6.1 K [62, 64]. Consistently with macroscopic magnetism, the system preserves the high magnetic field-Gd5 Ge4 structure, which is ferromagnetic, and whose concentration is only slightly reduced from ∼93% to ∼86% when the magnetic field is lowered isothermally from 3.5 T to 0. At 25 K, where the ferromagnetic–paramagnetic transformation becomes reversible, both the initial magnetization and the subsequent demagnetization paths are followed with high precision by the amount of the high magnetic field Gd5 Ge4 structure formed in the specimen, see Fig. 11.10 (left). It is worth noting that no more than 93 mol. % of the sample has been transformed from the low magnetic field Gd5 Ge4 to the high magnetic field Gd5 Ge4 structure by a 3.5 T field at either temperature, even though at this magnetic field the

11 Magnetocaloric Effect Associated with Magnetostructural Transitions

211

Fig. 11.10. Left – normalized magnetization (lines) and concentration of the high field Gd5 Ge4 structure (symbols) [63]; right – unit cell dimensions of the majority phase at 25 K [69]. Short straight arrows indicate the direction of the magnetic field change

magnetization is at ∼99% of its saturation value of 7.5 µB /Gd atom, indicating a nearly complete collinearity of the spin structure. The effect of the magnetic field on the crystal structure of Gd5 Ge4 is, therefore, different from the effect of temperature on the crystal structures of the related Gd5 Si2 Ge2 and Gd5 Si0.4 Ge3.6 phases. In the latter two systems, the respective temperature induced structural transitions are complete [39,45], yet a small but measurable fraction of Gd5 Ge4 retains its low field crystal structure while clearly ordering ferromagnetically. Even in a 3.5 T magnetic field, the crystallites of the ferromagnetic low field-Gd5 Ge4 (Sm5 Ge4 -type) remain large enough to be discernible in the powder diffraction data. As shown by Holm et al. [67], when Gd5 Ge4 is magnetized and subsequently demagnetized at 15 K, i.e., where the antiferromagnetic–ferromagnetic transformation is partially reversible [53, 61, 62, 65], the structural transformation between the Sm5 Ge4 - (low magnetic field) and Gd5 Si4 -types (high magnetic field) of crystal structure is also partially reversible. Thus, 48 mol% of the Gd5 Ge4 sample retains the high magnetic field crystal structure, while 52 mol% reversibly transforms into the low magnetic field polymorph after magnetizing and demagnetizing at 15 K [67]. According to Tang et al. [65], 49 mol% of the specimen retains the ferromagnetic state after magnetizing and demagnetizing at 15 K, while 51 mol% of the zero magnetic field antiferromagnetic state is recovered. The agreement between bulk magnetization and X-ray diffraction measurements, which were carried out independently, is excellent.

212

V.K. Pecharsky and K.A. Gschneidner, Jr.

Our primary concern here is the role of magnetostructural transitions in the appearance of the giant magnetocaloric effect, yet we should also emphasize the significant variations of the unit cell dimensions of Gd5 Ge4 , and therefore, predictable magnetic field induced shape changes of the material in either a single crystalline or in a textured polycrystalline states. The magnitude of the forced magnetostriction along the [100] direction places Gd5 Ge4 and other members of the R5 T4 family prominently among the so-called ferromagnetic shape memory alloys – materials in which considerable shape changes can be controlled by a varying magnetic field. As seen in Fig. 11.10 (right), the a-axis of Gd5 Ge4 reversibly contracts and expands by an enormous 1.9% (19, 000 parts per million) upon, respectively, magnetizing and demagnetizing the material. For comparison, the largest magnetic field induced shape changes known today are those observed in stressed Ni2 MnGa single crystals [71,72]. Up to 6% (theoretical) strains in Ni2 MnGa that could be induced by the application of a magnetic field along the [001] direction, occur as a result of rearranging twin variants in a ferromagnetic tetragonally distorted Heusler-type cubic structure. This rearrangement, which aligns the easy magnetization directions of all twin components with the external magnetic field vector, is however, not a true magnetic field-induced crystallographic phase transformation, as is the case in the R5 T4 materials. The ability of the magnetic field to induce a structural change in a material where spin–orbit contributions are negligible (Gd is an S-state ion), can be related to the coupling of the localized magnetic moments of the Gd atoms with the external magnetic field. The latter reverses the sign of magnetic exchange interactions and induces a metamagnetic transformation of the antiferromagnetic Gd5 Ge4 (negative exchange parameter) into the ferromagnetically ordered system (positive exchange parameter). The increasing magnetic field causes a spin flip and the resulting collinear or nearly collinear spin system minimizes its free energy via a martensitic magnetic field-induced structural transition from the low field-Gd5 Ge4 to the high field-Gd5 Ge4 structures. The sample remains both structurally [63, 67] and magnetically [61–65] inhomogeneous in a certain range of magnetic fields because the transition is magnetoelastic. Hence, work performed by the magnetic field to increase bulk magnetization is balanced by work required to overcome strain created by a 1.1% change of the phase volume. Since magnetic work (B dM ) becomes negligible near saturation as dM approaches zero but strain remains considerable, this may explain the presence of a small amount of the ferromagnetic low field-Gd5 Ge4 structure in a nearly completely magnetically saturated sample. The magnetocaloric effect of Gd5 Ge4 , calculated from magnetization data employing (11.3), is shown in Fig. 11.11. The isothermal magnetic entropy change of a virgin sample, i.e., when the material is in the antiferromagnetic state before the application of a magnetic field, is negligible as long as the field remains lower than the ∼1 T needed to induce the magnetostructural transition in Gd5 Ge4 . When Bf is 2 T and higher (Bi is 0 in all cases), the MCE rapidly increases in magnitude and extends towards the higher temperatures with increasing ∆B, consistent with the values of magnetic field

11 Magnetocaloric Effect Associated with Magnetostructural Transitions

213

Fig. 11.11. The MCE of Gd5 Ge4 calculated using (11.3). The plot on the left is obtained from M (H)T =const data measured during magnetizing a sample cooled in zero magnetic field from ∼150 K to each measurement temperature. The plot on the right is for M (H)T =const data measured after the same sample was magnetized by a 5 T field and then demagnetized

that are required to complete the magnetostructural transition in the compound [53, 61, 63, 65]. Conversely, when the sample was pre-magnetized by a 5 T magnetic field (the field, which is strong enough to induce the ferromagnetic state and complete the transformation below ∼30 K), an additional contribution to the MCE in the form of ∆SM peak centered around 17 K becomes clearly visible. Except for this additional contribution, the MCE in the pre-magnetized Gd5 Ge4 remains identical to that of the virgin sample. The behavior of the low temperature ∆SM peak in the pre-magnetized Gd5 Ge4 sample is consistent with that of a conventional ferromagnet, while the high-temperature component of ∆SM behaves as expected for a first-order phase transition material. Although quite unusual for a Gd-based compound, the dependence of the magnetocaloric effect on the magnetic history of the sample is in agreement with the existence of the ferromagnetic Gd5 Si4 -type structure induced either irreversibly (below ∼10 K) or partially reversibly (between ∼10 and ∼20 K) by a magnetic field. Recently, Holm et al. [69] showed that magnetic field also triggers a structural transformation in Gd5 (Si1.7 Ge2.3 ), which is another member of the same family of materials as Gd5 Ge4 , except that the former compound has the monoclinic Gd5 (Si2 Ge2 )-type crystal structure at room temperature. In the past, the structural transition to the orthorhombic Gd5 Si4 -type structure from the room temperature monoclinic phase of the Gd5 (Si4−x Gex ) system was assumed from temperature dependent X-ray diffraction data [39,43] and from the magnetic field dependent physical property measurements

214

V.K. Pecharsky and K.A. Gschneidner, Jr.

Fig. 11.12. Fragments of the X-ray powder diffraction patterns of Gd5 (Si1.7 Ge2.3 ) collected at 235 K with the magnetic field varying from 0 to 3.5 T and then back to 0 with a 0.5 T step

(e.g., see [40,43]). As illustrated in Fig. 11.12, in situ X-ray powder diffraction demonstrates that when a Gd5 (Si1.7 Ge2.3 ) sample is magnetized at T = 235 K, a structural transformation begins around 1.5 T and is nearly complete at 2.5 T. Upon subsequent isothermal demagnetization, the low magnetic field crystal structure begins to emerge at ∼1.5 T and the reverse transformation is complete at 0.5 T. Similar behavior is observed in Gd5 (Si1.7 Ge2.3 ) at other temperatures above its zero magnetic field TC . According to Holm et al. [69], the high magnetic field crystal structure of Gd5 (Si1.7 Ge2.3 ) is identical to the high magnetic field crystal structure of Gd5 Ge4 . In other words, both ferromagnetically ordered materials belong to the Gd5 Si4 -type, in which all interslab T–T bonds are present. This result also confirms that the high magnetic field polymorphs are the same as the low temperature zero magnetic field crystal structures of materials in this phase region of the Gd5 (Si4−x Gex ) system.

11.5 To What Extent a Structural Change Enhances the Giant Magnetocaloric Effect? In the previous section, we provided sufficient evidence that in the Gd5 (Si4−x Gex ) system, a martensitic-like structural transition follows ferromagnetic ordering/disordering when x  2 and that both the magnetic and crystallographic changes can be reversibly triggered by a magnetic field at

11 Magnetocaloric Effect Associated with Magnetostructural Transitions

215

ambient pressure. Although magnetic field-induced polymorphism apparently occurs in other giant magnetocaloric effect systems, such as the manganites [55], the Gd5 (Si4−x Gex ) alloys are unique because they enable one to obtain further insights on the role the structural transition plays in the enhancement of the magnetocaloric effect. This is especially true because alloys that have nearly identical stoichiometry, and therefore, closely spaced Curie temperatures around x = 2, can be prepared in two different crystal structures in the paramagnetic state [38, 41, 42]. When x is less than 1.91 [38], the paramagnetic Gd5 (Si4−x Gex ) alloys crystallize in the Gd5 Si4 -type and their crystal structures are preserved when they become ferromagnetic. As a result, the magnetic ordering in these materials occurs via a second-order phase transformation, i.e., Gd5 (Si4−x Gex ) alloys with x < 1.91 exhibit a conventional magnetocaloric effect, e.g., see Fig. 11.2. When x is 1.91 or greater, the paramagnetic Gd5 (Si4−x Gex ) phases adopt the Gd5 (Si2 Ge2 )-type structure and they order ferromagnetically concurrently with a structural change to the Gd5 Si4 -type structure (Fig. 11.12), which results in a first-order paramagnetic to ferromagnetic transformation and the giant magnetocaloric effect, see Figs. 11.3 and 11.11. Taking into consideration this difference in the thermodynamic nature of phase transitions occurring in closely related intermetallic phases, in Fig. 11.13 we illustrate the behavior of the total entropies of Gd5 (Si2.5 Ge1.5 ) and Gd5 (Si2 Ge2 ) in a zero magnetic field.3 The former orders ferromagnetically in a conventional (second-order) fashion at TC ∼ = 312 K, while the latter undergoes a first-order magnetostructural transition at TC ∼ = 270 K. Since the Curie temperatures of two materials are slightly different, all data in Fig. 11.13 are plotted as functions of the normalized temperature, T − Ttr , where Ttr is taken just above TC as the temperature at which the magnetostructural transformation in Gd5 (Si2 Ge2 ) is complete. For clarity, the total entropy functions of Gd5 (Si2.5 Ge1.5 ) were also normalized by subtracting a constant value of ∼5.5 J g-at−1 K−1 to match the total zero magnetic field entropy of Gd5 (Si2 Ge2 ) just below the magnetostructural transition. As expected from the similarity of the chemical compositions, identical crystallography and ferromagnetism, the total entropy functions of the two materials after normalization show nearly indistinguishable behavior in a zero magnetic field immediately below their respective Curie temperatures. When x ∼ = 2, either the monoclinic Gd5 (Si2 Ge2 )-type or the orthorhombic Gd5 Si4 type structures can be retained in the paramagnetic state at constant x by heat treating Gd5 (Si4−x Gex ) alloys at different temperatures [48, 51]. As we showed recently [73], it is conceivable, however, that the Gd5 Si4 -type structure in the paramagnetic state can be stabilized by oxygen as a result of heat treatment. Considering the martensitic-like mechanism of the structural transition (see Figs. 11.4 and 11.8) and the potential effect of oxygen, apparently located inside specific interslab voids [73], on the occurrence of the transformation, we will base our analysis on materials with close though different x but low and nearly the same amount of interstitial impurities, rather than on materials with the same x but different and usually uncontrolled impurity contents.

3

216

V.K. Pecharsky and K.A. Gschneidner, Jr.

Fig. 11.13. The behavior of the total entropies of Gd5 (Si2.5 Ge1.5 ), where ferromagnetic ordering at TC is a second-order phase transformation (solid lines), and of Gd5 (Si2 Ge2 ) where ferromagnetic ordering is a first-order phase transformation (dash-dotted lines) in the vicinities of their respective Curie temperatures in 0 and 7.5 T magnetic fields. The three vertical arrows around Ttr indicate the magnitudes of the magnetocaloric effects of Gd5 (Si2.5 Ge1.5 ) and Gd5 (Si2 Ge2 ), and entropy of the first-order phase transition in Gd5 (Si2 Ge2 ), all at T = Ttr

A major deviation occurs in the vicinity of TC , where the entropy of Gd5 (Si2 Ge2 ) is increased by ∆Sstr = ∆Hstr /Ttr , (∆Hstr is enthalpy or latent heat of a first-order phase transformation). It is worth noting that in theory, the total entropy is expected to undergo a discontinuous change by ∆Sstr at TC , but in reality, this change occurs over a few Kelvin wide temperature range. As is easy to see from Fig. 11.13, the two zero magnetic field entropy functions continue to behave very similarly above the TC , in fact, the difference between them remains nearly constant and equal to ∆Sstr . Once again, we relate this similarity in the behavior of the two entropies to close relationships between crystallography and magnetism of these two compounds. The total entropies of the two materials after the application of the magnetic field do not match as well as the zero magnetic field entropies do, as also clearly seen in Fig. 11.13. This mismatch can be related to different effects of the magnetic field on conventional ferromagnets, which Gd5 (Si2.5 Ge1.5 ) is, when compared to Gd5 (Si2 Ge2 ), which is the system with a first-order magnetic-martensitic phase change. In fact, since the crystal structure is not coupled to a magnetic sublattice in the case of Gd5 (Si2.5 Ge1.5 ), the magnetic field does not actually transform the system into a nearly collinear ferromagnet, nor does it raise the Curie temperature of the material.

11 Magnetocaloric Effect Associated with Magnetostructural Transitions

217

Conversely, in the case of Gd5 (Si2 Ge2 ), where the magnetic field induces a magnetostructural phase change, the Curie temperature is increased nearly linearly with field [26]. Furthermore, the conventional ferromagnetic ordering of the orthorhombic paramagnetic Gd5 (Si2 Ge2 ) would occur at a much higher temperature than a conventional ferromagnetic ordering of the monoclinic paramagnetic Gd5 (Si2 Ge2 ) provided the latter crystal lattice can co-exist with a ferromagnetic sublattice [51]. This difference in Curie temperatures of the two structures nearly completely eliminates spin fluctuations and magnetic clustering in the case of Gd5 (Si2 Ge2 ), while both should be present in Gd5 (Si2.5 Ge1.5 ). The magnetocaloric effects at TC are shown in Fig. 11.13 as arrows for both materials (the arrows are offset along the temperature axis for clarity). Obviously, the two MCE’s are considerably different and the difference between them must be ascribed to the absence and the presence of a structural change in Gd5 (Si2.5 Ge1.5 ) and Gd5 (Si2 Ge2 ), respectively. Even though the magnetic field has slightly different effect on the total entropies of the two compounds, the difference in ∆SM at TC is nearly identical to ∆Sstr . The large magnetic field induced phase volume and chemical bonding changes, observed in Gd5 Ge4 , Gd5 (Si1.7 Ge2.3 ), and other R5 T4 materials, therefore, indicate that the giant MCE is achieved due to the concomitant change of the entropy during the structural transformation, designated in Fig. 11.13 as ∆Sstr . As a result, it is possible to speculate that the observed giant magnetocaloric effect is the sum of the conventional magnetic entropy-driven process (11.3) and the difference in the entropies of the two crystallographic modifications (∆Sstr ) of a solid   Bf  ∂M (T, B) ∆SM (T, ∆B)∆B=Bf −Bi = ∆Sstr + dB . (11.5) ∂T Bi B Although the first factor in the right-hand side of (11.5) is a hidden parameter in conventional magnetization, heat capacity and direct magnetocaloric effect measurements because either property reflects both the magnetic and crystallographic states of the material [e.g., imagine the absence of the data for Gd5 (Si2.5 Ge1.5 ) in Fig. 11.13], an estimate based on comparing the MCE’s exhibited by these two members of the Gd5 (Si4−x Gex ) family is possible. This is illustrated in Fig. 11.14, where we plot the isothermal magnetic entropy change of Gd5 (Si2 Ge2 ) taken from [29] and the same of Gd5 (Si2.5 Ge1.5 ) [74] for ∆B varying from 2 to 10 T. Considering (11.5) and recalling that ∆Sstr is independent of the magnetic field, provided the entire volume of the material retains the low magnetic field crystal structure at Bi and it is completely converted into the high magnetic field polymorph by Bf , the difference between the peak values of ∆SM should remain constant (or nearly constant in view of Fig. 11.13) regardless of ∆B. This constant difference, indeed, is approximately equal to ∆Sstr . Numerical data representing the peak ∆SM values for both materials are listed in Table 11.1. After recalling that the crystallographic phase change

218

V.K. Pecharsky and K.A. Gschneidner, Jr.

Fig. 11.14. The magnetocaloric effects of Gd5 (Si2 Ge2 ) (left) and Gd5(Si2.5 Ge1.5 ) (right) calculated from heat capacity measured in constant magnetic fields of 0, 2, 5, 7.5, and 10 T. In every case, Bi = 0 Table 11.1. Peak values of the magnetocaloric effect of Gd5 (Si2 Ge2 ) and Gd5 (Si2.5 Ge1.5 ) for different ∆B (Bi is always 0) and ∆Sstr estimated as the average difference between the peak ∆SM values of the two compounds Compound

Gd5 (Si2 Ge2 ) Gd5 (Si2.5 Ge1.5 ) Difference Average

Isothermal magnetic entropy change, −∆SM (J g-at−1 K−1 ) ∆B = 2 T

∆B = 5 T

1.75 0.57 1.19 (excluded)

2.15 1.03 1.12

∆B = 7.5 T 2.42 1.33 1.09 ∆Sstr = 1.08(4)

∆B = 10 T 2.61 1.57 1.04

was not complete even in a 3.5 T magnetic field in Gd5 Ge4 and assuming that similar effect is feasible in Gd5 (Si2 Ge2 ), it only makes sense to discard ∆SM difference for the smallest ∆B, where Bf = 2 T. The values for the remaining three field changes are constant to within one standard deviation of the mean, which is ∆Sstr = 1.08(4) J g-at−1 K−1 . The latter is of the same order of magnitude as the entropies of transformations of pure metals, which could be approximated [75] as ∆Sstr = 0.75ν J g-at−1 K−1 for an fcc → bcc transition (the actual values vary from 0.5 to 3.01 J g-at−1 K−1 for γ → δ transitions in Fe and Ce, respectively), or as ∆Sstr = 0.62 + 0.75ν J g-at−1 K−1 for a hcp → bcc transition (the actual values vary from 0.56 J g-at−1 K−1 in Li to 3.68 J g-at−1 K−1 in Ti), where ν is the number of valence electrons. Analysis of data presented in Fig. 11.14 and Table 11.1 indicates that ∆Sstr accounts for 50% or more of the total extensive magnetocaloric effect

11 Magnetocaloric Effect Associated with Magnetostructural Transitions

219

in magnetic fields below 5 T. Its relative contribution to ∆SM increases as the magnetic field decreases as long as the final magnetic field (Bf ) is strong enough to complete the magnetostructural transition. Advanced magnetocaloric materials should exist in other solid systems where structural changes are coupled with ferromagnetic ordering, and therefore, can be triggered by a magnetic field. Considering (11.5), the strongest MCEs in the weakest magnetic fields should be found in novel materials, engineered in order to maximize the difference in the entropies of the low magnetic field and high magnetic field phases, and therefore, the resulting entropy of a structural transformation, ∆Sstr .

11.6 Conclusions In summary, analysis of the magnetic-martensitic transformations and the magnetocaloric effects in a few representatives of the R5 T4 family of intermetallic compounds – Gd5 Ge4 , Gd5 (Si1.7 Ge2.3 ), Gd5 (Si2 Ge2 ), and Gd5 (Si2.5 Ge1.5 ) – indicates that the coupling between the magnetic and crystal lattices is critical in bringing about the giant magnetocaloric effect. Although in general it is difficult to separate the contributions from either sublattice into the MCE, in the R5 T4 series it was possible experimentally because these materials can be prepared in two different states: one that exhibits a conventional MCE and another where the magnetic and crystal lattices are coupled and the giant magnetocaloric effect is the result of a first-order magnetostructural transformation. Especially the extensive measure of the magnetocaloric effect, i.e., the isothermal magnetic field-induced entropy change, can be strongly enhanced by the added entropy of a structural transition provided the magnetic field completes the polymorphic transformation. This additional entropy is magnetic field independent, and therefore, may account for more than a half of the observed magnetocaloric effect, especially in relatively weak magnetic fields. Design of novel materials exhibiting potent magnetocaloric effects in magnetic fields that can be created by using modern permanent magnets is, therefore, possible by ensuring that (i) a structural transition can be triggered and completed or nearly completed by a weak magnetic field; (ii) the transformation has low thermal and magnetic field hystereses, i.e., is reversible, and (iii) the difference between the total entropies of two polymorphic modifications, and therefore, ∆Sstr , is maximized. Acknowledgements We thank Dr. A.P. Holm for allowing us to use some of his unpublished temperature and magnetic field dependent X-ray powder diffraction data and to Dr. A.O. Pecharsky for allowing us to use some of her unpublished heat capacity data collected as functions of temperature in magnetic fields ranging from

220

V.K. Pecharsky and K.A. Gschneidner, Jr.

0 to 10 T. This work was supported by the Office of Basic Energy Sciences, Materials Sciences Division of the United States Department of Energy under Contract No. W-7405-ENG-82 with the Iowa State University of Science of Technology.

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18.

19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29.

E. Warburg: Ann. Phys. (Leipzig) 13, 141 (1881) P. Debye: Ann. Phys. (Leipzig) 81, 1154 (1926) P. Weiss, R. Forrer: Ann. Phys. (Leipzig) 5, 153 (1926) W.F. Giauque: J. Am. Chem. Soc. 49, 1864 (1927) P.J. von Ranke, V.K. Pecharsky, K.A. Gschneidner Jr.: Phys. Rev. B 58, 12 110 (1998) A.M. Tishin, K.A. Gschneidner Jr., V.K. Pecharsky: Phys. Rev. B 59, 503 (1999) V.K. Pecharsky, K.A. Gschneidner Jr., A.O. Pecharsky, A.M. Tishin: Phys. Rev. B 64, 144 406/1 (2001) A.L. Lima, A.S. Oliveira, A.M. Gomes, P.J. von Ranke: Phys. Rev. B 65, 172 411/1 (2002) H. Yamada, T. Goto: Phys. Rev. B 68, 184 417/1 (2003) P.J. von Ranke, N.A. de Oliveira, S. Gama: Phys. Let. A 320, 302 (2003) A.M. Tishin, Y.I. Spichkin: The magnetocaloric effect and its applications (IOP, Bristol Philadelphia 2003) W.F. Giauque, D.P. MacDougall: Phys. Rev. 43, 768 (1933) G.V. Brown: J. Appl. Phys. 47, 3673 (1976) C.B. Zimm, A. Jastrab, A. Sternberg, V.K. Pecharsky, K.A. Gschneidner Jr., M.G. Osborne, I.E. Anderson: Adv. Cryo. Engin. 43, 1759 (1998) K.A. Gschneidner Jr., V.K. Pecharsky, C.B. Zimm: Mater. Techn. 12, 145 (1997) V.K. Pecharsky, K.A. Gschneidner Jr.: J. Magn. Magn. Mater. 200, 44 (1999) K.A. Gschneidner Jr., V.K. Pecharsky: Ann. Rev. Mater. Sci. 30, 387 (2000) C.B. Zimm, A. Sternberg, A.G. Jastrab, A.M. Boeder, L.M. Lawton Jr., J.J. Chell: Rotating bed magnetic refrigeration apparatus. US Patent No. 6,526,759 (2003) C.B. Zimm, A. Sternberg, A.G. Jastrab, A.M. Boeder, L.M. Lawton Jr.: Rotating magnet magnetic refrigerator. US Patent No. 6,668,560 (2003) V.K. Pecharsky, K.A. Gschneidner Jr.: J. Appl. Phys. 86, 565 (1999) B.R. Gopal, R. Chahine, M. F¨ old`eaki, T.K. Bose: Rev. Sci. Instrum. 66, 232 (1995) B.R. Gopal, R. Chahine, T.K. Bose: Rev. Sci. Instrum. 68, 1818 (1997) S.Y. Dan’kov, A.M. Tishin, V.K. Pecharsky, K.A. Gschneidner Jr.: Rev. Sci. Instrum. 68, 2432 (1997) A.H. Morrish: The physical principles of magnetism (Wiley, New York 1965) V.N. Belova, V.I. Nikolaev, V.M. Stuchebnikov: Fiz. Tv. Tela. (Leningrad) 15, 909 (1973) V.K. Pecharsky, K.A. Gschneidner Jr.: Phys. Rev. Lett. 78, 4494 (1997) V.K. Pecharsky, K.A. Gschneidner Jr.: Appl. Phys. Lett. 70, 3299 (1997) V.K. Pecharsky, K.A. Gschneidner Jr.: J. Magn. Magn. Mater. 167, L179 (1997) A.O. Pecharsky, K.A. Gschneidner Jr., V.K. Pecharsky: J. Appl. Phys. 93, 4722 (2003)

11 Magnetocaloric Effect Associated with Magnetostructural Transitions

221

30. K.A. Gschneidner Jr., V.K. Pecharsky, E. Br¨ uck, H.G.M. Duijn, E.M. Levin: Phys. Rev. Lett. 85, 4190 (2000) 31. S.A. Nikitin, G. Myalikgulev, A.M. Tishin, M.P. Annaorazov, K.A. Asatryan, A.L. Tyurin: Phys. Lett. A 148, 363 (1990) 32. M.P. Annaorazov, S.A. Nikitin, A.L. Tyurin, K.A. Asatryan, A.K. Dovletov: J. Appl. Phys. 79, 1689 (1996) 33. J.F. Herbst, C.D. Fuerst, R.D. McMichael: J. Appl. Phys. 79, 5998 (1996) 34. Z.B. Guo, Y.W. Du, J.S. Zhu, H. Huang, W.P. Ding, D. Feng: Phys. Rev. Lett. 78, 1142 (1997) 35. G.S. Smith, A.G. Tharp, Q. Johnson: Nature (London) 210, 1148 (1966) 36. G.S. Smith, A.G. Tharp, Q. Johnson: Act. Cryst. 22, 940 (1967) 37. F. Holtzberg, R.J. Gambino, T.R. McGuire: J. Phys. Chem. Solids 28, 2283 (1967) 38. V.K. Pecharsky, K.A. Gschneidner Jr.: J. Alloys Compds. 260, 98 (1997) 39. W. Choe, V.K. Pecharsky, A.O. Pecharsky, K.A. Gschneidner Jr., V.G. Young Jr., G.J. Miller: Phys. Rev. Lett. 84, 4617 (2000) 40. V.K. Pecharsky, K.A. Gschneidner Jr.: Adv. Mater. 13, 683 (2001) 41. A.O. Pecharsky, K.A. Gschneidner Jr., V.K. Pecharsky: J. Magn. Magn. Mater. 267, 60 (2003) 42. A.O. Pecharsky, K.A. Gschneidner Jr., V.K. Pecharsky, C.E. Schindler: J. Alloys Compds. 338, 126 (2002) 43. L. Morellon, P.A. Algarabel, M.R. Ibarra, J. Blasco, B. Garc´ıa-Landa, Z. Arnold, F. Albertini: Phys. Rev. B 58, R14 721 (1998) 44. L. Morellon, J. Stankiewicz, B. Garc´ıa-Landa, P.A. Algarabel, M.R. Ibarra: Appl. Phys. Lett. 73, 3462 (1998) 45. L. Morellon, J. Blasco, P.A. Algarabel, M.R. Ibarra: Phys. Rev. B 62, 1022 (2000) 46. Q.L. Liu, G.H. Rao, H.F. Yang, J.K. Liang: J. Alloys Compds. 325, 50 (2001) 47. Q.L. Liu, G.H. Rao, J.K. Liang: Rigaku J. 18, 46 (2001) 48. V.K. Pecharsky, A.O. Pecharsky, K.A. Gschneidner Jr.: J. Alloys Compds. 344, 362 (2002) 49. M. Nazih, A. de Visser, L. Zhang, O. Tegus, E. Br¨ uck: Solid State Comm. 126, 255 (2003) 50. M. Han, D.C. Jiles, J.E. Snyder, C.C.H. Lo, J.S. Leib, J.A. Paulsen, A.O. Pecharsky: J. Appl. Phys. 93, 8486 (2003) 51. V.K. Pecharsky, G.D. Samolyuk, V.P. Antropov, A.O. Pecharsky, K.A. Gschneidner Jr.: J. Solid State Chem. 171, 57 (2003) 52. J.S. Meyers, L.S. Chumbley, F. Laabs, A.O. Pecharsky: Acta Mater. 51, 61 (2003) 53. C. Magen, L. Morellon, P.A. Algarabel, C. Marquina, M.R. Ibarra: J. Phys.: Cond. Mater. 15, 2389 (2003) 54. V.K. Pecharsky, K.A. Gschneidner Jr., A.O. Pecharsky, A.M. Tishin: Phys. Rev. B 64, 1444 406/1 (2001) 55. Z.M. Wang, G. Ni, Q.Y. Xu, H. Sang, Y.W. Du: J. Appl. Phys. 90, 5689 (2001) 56. H. Wada, Y. Tanabe: Appl. Phys. Lett. 79, 3302 (2001) 57. J.P. Meyer, P. Taglang: Compt. Rend. 264, 1820 (1958) 58. P.I. Kripyakevich, O.S. Zarechnyuk, E.I. Gladyshevskii, O.I. Bodak: Z. Anorg. Allg. Chem. 358, 90 (1968) 59. O. Tegus, E. Br¨ uck, K.H.J. Buschow, F.R. de Boer: Nature (London) 415, 150 (2002)

222

V.K. Pecharsky and K.A. Gschneidner, Jr.

60. X.X. Zhang, G.H. Wen, F.W. Wang, W.H. Wang, C.H. Yu, G.H. Wu: Appl. Phys. Lett. 77, 3072 (2000) 61. E.M. Levin, V.K. Pecharsky, K.A. Gschneidner Jr., G.J. Miller: Phys. Rev. B 64, 235 103/1 (2001) 62. E.M. Levin, K.A. Gschneidner Jr., V.K. Pecharsky: Phys. Rev. B 65, 214 427/1 (2002) 63. V.K. Pecharsky, A.P. Holm, K.A. Gschneidner Jr., R. Rink: Phys. Rev. Lett. 91, 197 204/1 (2003) 64. C. Magen, Z. Arnold, L. Morellon, Y. Skorokhod, P.A. Algarabel, M.R. Ibarra, J. Kamarad: Phys. Rev. Lett. 91, 207 202/1 (2003) 65. H. Tang, V.K. Pecharsky, K.A. Gschneidner Jr., A.O. Pecharsky: Phys. Rev. B 65, 064 410/1 (2004) 66. V. Hardy, S. Majumdar, S.J. Crowe, M.R. Lees, D. McK. Paul, L. Herv´e, A. Maignan, S. H´ebert, C. Martin, C. Yaicle, M. Hervieu, B. Raveau: Phys. Rev. B 69, 020 407/1 (2004) 67. A.P. Holm, V.K. Pecharsky, K.A. Gschneidner Jr., R. Rink, M.N. Jirmanus: Rev. Sci. Instr. 75, 1081 (2004) 68. E.M. Levin, K.A. Gschneidner Jr., T.A. Lograsso, D.L. Schlagel, V.K. Pecharsky: Phys. Rev. B 69, 144 428/1 (2004) 69. A.P. Holm, V.K. Pecharsky, K.A. Gschneidner Jr.: (2003) unpublished 70. Numerical zero magnetic field linear thermal expansion data of polycrystalline Gd5 Ge4 [53] measured using a strain gauge that were employed to compute the linear thermal expansion coefficient were kindly provided by Dr. L. Morellon 71. K. Ullakko, J.K. Huang, C. Kantner, R.C. O’Handley, V. Kokorin: Appl. Phys. Lett. 69, 1966 (1996) 72. S.J. Murray, M. Marioni, S.M. Allen, R.C. O’Handley, T.A. Lograsso: Appl. Phys. Lett. 77, 886 (2000) 73. Y. Mozharivskyj, A.O. Pecharsky, V.K. Pecharsky, G.J. Miller: J. Am. Chem. Soc. 127, 317 (2005) 74. A.O. Pecharsky, K.A. Gschneidner Jr., V.K. Pecharsky: (2004) unpublished 75. K.A. Gschneidner Jr.: J. Less Common Metals 43, 179 (1975)

12 Entropy Change and Magnetocaloric Effect in Magnetostructural Transformations F. Casanova, X. Batlle, A. Labarta, J. Marcos, E. Vives, and L. Ma˜ nosa, and A. Planes

12.1 Introduction The physics of the magnetocaloric effect (MCE) is an old subject with renewed interest in the last decades owing to its potential application for refrigeration [1]. In fact, magnetic refrigeration offers the prospect of an energy-efficient and environment friendly alternative to the common vapor-cycle refrigeration technology in use today [2, 3]. The MCE is commonly defined as the heating or cooling (i.e., the temperature change) of a magnetic material due to the application of a magnetic field. This effect has been called adiabatic demagnetisation for years, though this phenomenon is one particular practical application of the MCE in magnetic materials. For excellent reviews on the magnetocaloric effect, see [1, 4, 5]. All magnetic materials exhibit MCE, which is an intrinsic phenomenon, although the intensity of this effect shows a large variation depending on the material. MCE originates from the coupling between the magnetic sublattice and the applied magnetic field, H, which produces an additional contribution to the entropy of the solid. Thus, the strength of the MCE in a particular solid can be characterised as the isothermal entropy change, ∆Siso , or the adiabatic temperature change, ∆Tad , arising from the application or removal of an external magnetic field. In particular, ∆Siso induced in a system when the magnetic field is varied from Hi to Hf can be experimentally determined directly from calorimetric measurements or indirectly from magnetisation curves by numerical integration of the Maxwell relation:   Hf  ∂M dH . (12.1) ∆Siso (Hi → Hf ) = ∂T H Hi According to this equation, the absolute value of the derivative of magnetisation with respect to temperature should be large to get a large entropy change. This occurs, for instance, when the system undergoes a magnetic transition. However, in second-order magnetic transitions, the existence of short-range

224

F. Casanova et al.

order and spin fluctuations above the ordering temperature brings about a reduction in the maximum value of (∂M/∂T )H , and consequently the maximum MCE is reduced. In contrast, an ideal first-order phase transition takes place at constant temperature, so the value of (∂M/∂T )H should be infinitely large. Actually, in an ideal first-order magnetic transition, the discontinuities in both magnetisation and entropy make it possible that (12.1) can be rewritten in terms of the Clausius–Clapeyron equation as ∆Siso = −∆M

dHt , dTt

(12.2)

where Ht is the field at which the transition takes place and Tt is the transition temperature. Therefore, the MCE may be maximised in the vicinity of a first-order structural transition, when a large entropy change related to the crystallographic transformation, ∆S, can be field-induced through magnetoelastic coupling. The existence of a large ∆S associated with the first-order transition brings about an extra contribution to the MCE, yielding the socalled giant MCE. The use of this ∆S in practical applications is possible provided that the phase transition can be induced by an external magnetic field [6]. Extensive search for materials undergoing simultaneously structural and magnetic transitions (magnetostructural transformations) has recently been described in literature [7–9]. One of the most promising class of materials displaying a magnetostructural transformation comprises Gd5 (Six Ge1−x )4 alloys with 0 ≤ x ≤ 0.5 [7, 10], in which ∆Siso values at least twice larger than that of Gd near room temperature for H = 50 kOe are observed. This enhanced MCE is due to a strong interplay between the magnetic and structural features in this system. These alloys have crystallographic structures made up of parallel nanometric slabs connected by covalent-like bonds [11]. Interestingly, the interslab connectivity does not only depend on x and the actual crystallographic structure, but also on the magnetic ordering. This makes it possible to field-induce a structural transformation in the system, driven by a shear mechanism between the parallel slabs, which adapts the interslab connectivity to the final magnetic state. Therefore, in these alloys, the magnetoelastic coupling responsible for the magnetostructural transformation is operating at the microscopic scale of the nanometric slabs. However, in other materials undergoing magnetostructural transformations, magnetoelastic coupling may occur at multiple scales ranging from microscopic to mesoscopic lengths. Moreover, competing mechanisms of magnetoelastic coupling taking place at different scales may be present simultaneously. This is the case in a series of magnetic composition-related Ni–Mn–Ga shape-memory alloys undergoing a martensitic transformation. It is shown that the MCE, increasing the magnetic field in the vicinity of the martensitic transition, mainly originates from two contributions with opposite sign [12]: (i) magnetoelastic coupling on mesoscopic scale between the magnetic domains and the martensitic variants which gives a positive contribution to the MCE, and (ii) the microscopic spin–phonon coupling which produces a shift of Tt

12 Entropy Change and Magnetocaloric Effect

225

with H and gives a negative contribution to the MCE. The relative importance of these two contributions has been shown to vary with composition. It is worth noting that, in the range of composition for which the mesoscopic mechanism becomes dominant, (∂M/∂T )H may be positive in the vicinity of the martensitic transformation when the magnetic field is increased. This is a very unusual behaviour for a magnetic system. In this chapter, the magnetostructural transformations and the MCE exhibited by both families of alloys (Gd–Si–Ge and Ni–Mn–Ga) are studied, with a discussion of the different scales at which magnetoelastic coupling takes place.

12.2 Multiscale Origin of the MCE in Ni–Mn–Ga Alloys Ni–Mn–Ga alloys are materials with twofold interest, as they display both magnetocaloric and magnetic shape-memory effects. The magnetic shapememory property is a consequence of the coupling between structural and magnetic degrees of freedom in a ferromagnetic alloy undergoing a martensitic transformation at a temperature, TM , below the Curie temperature (TC ). Ferromagnetic Ni–Mn–Ga alloys close to the stoichiometric Heusler composition are prototypical materials exhibiting this property, in which martensitic transformation occurs in the ferromagnetic state from a cubic structure (parent phase) to a tetragonal martensitic phase below TM . One important feature to understand for the magnetic behaviour of this system is that, in the tetragonal phase, there are twin related structural domains with strong uniaxial anisotropy compared to that of the cubic phase, which can be rearranged by the application of moderate fields. Consequently, the martensitic transformation takes place between two ferromagnetic states with very different magnetic domain structures, being accompanied by a jump in the magnetisation of the system, which is mainly controlled by the magnetoelastic coupling at the mesoscopic scale of the martensitic variants. For certain compositions, this magnetisation jump may produce at moderate fields an MCE larger than that observed in Gd for a similar range of temperatures. We present results for a single crystal of composition Ni49.5 Mn25.4 Ga25.1 . The sample exhibited an L21 structure at room temperature, a TC of about 381 K and a martensitic transition on cooling at TM ≈ 177 K to a modulated tetragonal structure (5R). The hysteresis for the martensitic transition was ∼10 K. Isothermal magnetisation measurements were carried out in this specimen by the extraction technique in the range from 150 to 200 K, under magnetic fields up to 40 kOe. Figure 12.1a shows the obtained curves for selected temperatures, from which the field-induced entropy change, when the field increases from 0 to H, can be evaluated by numerical integration of (12.1). The results are plotted in Fig. 12.1b for selected values of H. The two peaks observed in the temperature dependence of ∆Siso (0 → H) originate from unavoidable small composition inhomogeneities and impurities, which

226

F. Casanova et al.

Fig. 12.1. (a) Magnetisation versus applied field at different temperatures around TM for a Ni49.5 Mn25.4 Ga25.1 sample. (b) Field-induced entropy change as a function of temperature for selected values of the maximum applied field, calculated from the data in (a)

Fig. 12.2. Average field-induced entropy change ∆S as a function of H for a Ni49.5 Mn25.4 Ga25.1 sample. The continuous line is a fit of the experimental data to a phenomenogical model, which includes two contributions to the magnetoelastic coupling: the spin–phonon microscopic coupling (dotted line) and the martensitevariant mesoscopic coupling (dashed line)

are known to slightly affect the actual transition path [13]. In order to compare ∆Siso (0 → H, T ) for different samples it is convenient to calculate, for each field, the average of the field-induced entropy change over ∆T (H) (the temperature range at which the transition spreads) as  1 ∆Siso (0 → H, T ) dT . (12.3) ∆S(H) = ∆T (H) ∆T This integral is performed numerically by taking a suitable baseline which permits elimination of any contribution arising from temperature variations of the magnetisation outside the transformation region. The obtained ∆S(H) values as a function of the applied magnetic field are plotted in Fig. 12.2.

12 Entropy Change and Magnetocaloric Effect

227

The average entropy change first increases, reaches a maximum at 13 kOe, and decreases linearly with H at higher fields. However, the most remarkable fact in these data is the positive value of entropy change originating from an increase in the applied field. A deeper understanding of the magnetoelastic coupling in this system, which is at the origin of this peculiar magnetic behaviour, can be gained by analysing the temperature dependence of the magnetisation curves measured at a fixed value of the magnetic field around TM , which are displayed in Fig. 12.3a. In the transition region, a jump in the magnetisation isofield curves is observed, which is positive or negative depending on the strength of the applied magnetic field. The magnitude of this jump ∆M (magnetisation difference between martensite and parent phase) displays a strong dependence on H, as shown in Fig. 12.3b. The different regions can be understood by taking into account the magnetic and structural domain patterns formed as a function of H [14, 15]. When the martensitic transformation takes place at zero field, nucleation gives rise to plates formed by parallel stripes of twin related variants with strong uniaxial anisotropy. The nucleation of the plates with such a microstructure is very advantageous, since it minimises the elastic (strain) energy arising from the crystal lattice misfit along the interfacial boundaries. At low fields, magnetic domains are formed within each variant, in such a way that the magnetisation alternates between two values, which are parallel to the corresponding easy axes. This configuration ensures the absence of magnetic poles in the martensite phase giving rise to negative values of ∆M .

Fig. 12.3. (a) Magnetisation versus temperature curves at selected values of the magnetic field applied along the [100] direction of the cubic phase. Measurements were performed during heating. The different symbols correspond to the following fields: H = 0 (filled square), 100 Oe (filled circle), 200 Oe (filled triangle), 300 Oe (filled down triangle), 400 Oe (filled diamond ), 500 Oe (filled triangle left), 700 Oe (filled triangle right), 1 kOe (open square), 2 kOe (open circle), 4 kOe (open triangle), 6 kOe(open down triangle), 10 kOe (open diamond ), 20 kOe (open triangle left) and 50 kOe (open triangle right). (b) Magnetisation change at the phase transition, as a function of the magnetic field

228

F. Casanova et al.

In contrast, if the sample is cooled through the transition under a magnetic field larger than the saturation field of the parent phase, the twin related variants are magnetised. As the field is further increased, due to the strong uniaxial anisotropy of the tetragonal phase, the Zeeman energy is minimised by increasing the fraction of those variants with the easy magnetisation axis forming a smaller angle with the applied field. At high enough fields, transformation results in magnetically saturated single variant crystal, and ∆M becomes positive and constant. Interestingly, this magnetic behaviour is controlled by the same magnetoelastic coupling which causes the magnetic shapememory effect [16, 17]. Consequently, ∆M and the field-induced entropy change through the transition are mainly controlled by the magnetoelastic coupling at the mesoscopic scale of the twin related variants. This contribution to the entropy change is positive and dominant at low fields. In fact, for moderate values of H, the field-induced entropy is not related to the entropy change of the structural transition. This is because of the weak spin–phonon coupling at the microscopic scale, as proven by the weak dependence of the phonon-dispersion curves and TM on H [18, 19]. High fields are required to induce the transition even at temperatures very close to TM . However, the magnetoelastic coupling at the microscopic scale is not entirely negligible in this material, giving rise to a negative contribution to the entropy change which becomes dominant at high fields and in turn gives rise to the linear decrease of ∆S(H) shown in Fig. 12.2 above 13 kOe. A simple phenomenological model including the interplay between both scales of magnetoelastic coupling accounts well for the observed experimental behaviour of ∆S(H) [12], as shown in Fig. 12.2. In this model, the contribution coming from the magnetoelastic coupling at the mesoscopic scale of the martensite variants is evaluated assuming that, for the sake of simplicity, the system contains only two twin-related variants, whose relative population varies exponentially as H increases above the saturation field of the parent phase. The contribution arising from the spin–phonon coupling is calculated from (12.2), taking (dTM /dH)H as a measure of the shift of TM induced by the field. Both contributions are depicted separately in Fig. 12.2 to emphasise their different field dependence. Thus, the mesoscopic contribution is positive and increases monotonically to reach saturation, while the microscopic contribution is negative and decreases linearly as H increases. The relative importance of these two contributions changes with the actual composition of the alloy. In Fig. 12.4, ∆S(H) obtained from magnetisation data [12] for five alloys of different compositions, including that of Fig. 12.2, are depicted. The continuous lines are the fits to the model of [12]. In alloys with a large difference between TC and TM (alloys 1 and 2 in Fig. 12.4), mesoscopic coupling is dominant and a negative giant MCE (decrease in temperature caused by an adiabatic increase of the magnetic field) is induced at moderate fields. In contrast, for alloys in which these temperatures are very close to one another (case 5 in Fig. 12.4), the microscopic coupling is the most relevant contribution

12 Entropy Change and Magnetocaloric Effect

229

Fig. 12.4. Average field-induced entropy change ∆S as a function of H for all the different analysed alloys. The continuous lines are the fits of the experimental data to a phenomenogical model

and gives rise to a positive giant magnetocaloric effect. This is in agreement with the decrease of the anisotropy of martensite phase as TM approaches TC [20]. Therefore, the maximum values of the field-induced entropy change in Ni–Mn–Ga alloys correspond to those with TM ≈ TC and are approximately −30 J kg−1 K−1 for a field of 15 kOe [21]. These values are even larger than maximum values reported for prototypical materials displaying giant MCE at room temperature such as Gd–Si–Ge [22] and Mn–As [9] (of the order of −15 J kg−1 K−1 for a field of 20 kOe).

12.3 Direct Determination of the Entropy Change at a First-Order Transformation As has been discussed in Sect. 12.1, the MCE in a given magnetic material undergoing a first-order structural transition can be characterised by quantifying the field-induced ∆Siso indirectly from magnetisation data, as it is the case for Ni–Mn–Ga in Sect. 12.2, or directly from calorimetric measurements. Differential scanning calorimeters (DSC) are well adapted to determine ∆S at a first-order transformation, since with this technique the heat flow exchanged by the sample and the surroundings is directly measured and a proper integration of the calibrated signal yields the latent heat of the transformation. A DSC operating under magnetic field is particularly well suited to study magnetostructural transitions [23]. In these devices, the temperature is con˙ exchanged between tinuously scanned at a rate T˙ , while the heat flow, Q, the sample and the thermal block is measured by the thermobatteries. From ˙ the derivative of the heat with respect to the temperature can be T˙ and Q,

230

F. Casanova et al.

obtained (calorimetric signal). A proper integration of this signal subtracting the baseline leads to the latent heat or the entropy change of the transition. It is worth noting that the calorimeter can also operate by sweeping the field maintaining a constant temperature. In this mode of operation, a magnetostructural transformation can be field-induced by setting a temperature above but sufficiently close to Tt . In this case, the calorimetric signal can be obtained from Q˙ and H˙ and a proper integration of this signal provides the field-induced entropy change. Consequently, this system permits a direct determination of the MCE, since ∆S induced in the sample by the application of an external magnetic field can be directly determined.

12.4 Magnetostructural Transformation in Gd–Si–Ge Alloys Gd5 (Six Ge1−x )4 alloys are one of the most promising materials for practical applications of the MCE, which in these materials originates from a first-order magnetostructural transformation that appears in two compositional ranges. The phase diagram is complex [22, 24, 25], but the most relevant feature pertinent to the MCE is the existence of a structural transformation which takes place simultaneously with a magnetic transition from a high-temperature paramagnetic (PM) monoclinic phase to a low-temperature ferromagnetic (FM) Gd5 Si4 -type orthorhombic phase for 0.24 ≤ x ≤ 0.5 [22, 24] and from a high temperature antiferromagnetic (AFM) Gd5 Ge4 -type orthorhombic phase to an isostructural low-temperature FM phase for x ≤ 0.2 [25]. At higher temperature, a second-order transition occurs from the AFM to the PM state for x ≤ 0.2. The giant MCE in these alloys is due to the existence of a strong coupling between the structure and the magnetic ordering which is governed by the interslab connectivity through the number of covalent-like bonds [11]. In this section, we present a direct determination of ∆S at the first-order transformation in Gd5 (Six Ge1−x )4 alloys. The samples were prepared by arcmelting and heat-treated for 9 h at 920◦ C under 10−5 mbar. Figure 12.5 shows an example of the measured calorimetric curves by sweeping the temperature on cooling under different values of the magnetic field for samples with x = 0.1 (left panel) and x = 0.3 (right panel). The magnetostructural transformation corresponds to the sharp peak, which is present both for x = 0.1 and 0.3, and shifts significantly with the applied magnetic field, revealing a strong magnetoelastic coupling. The second-order transition appears only for samples with x ≤ 0.2 as a small lambda peak, which shows weak field dependence. At high enough fields, the first- and second-order transitions collapse and the structural transformation takes place between the FM and PM states. It is worth noting that the transition temperature of the structural transformation for moderate fields, lower than that at which the first- and secondorder transitions collapse, varies linearly with the applied field. The reciprocal

12 Entropy Change and Magnetocaloric Effect

231

Fig. 12.5. Calorimetric curves recorded on cooling through the transitions of Gd5 (Six Ge1−x )4 samples (x = 0.1 and 0.3) for different applied magnetic fields. Insets show the entropy change as a function of magnetic field

of the slope of this linear dependence defined as dHt /dTt (Ht is the transition field, i.e., the field at which the transition takes place at Tt ) is displayed in Fig. 12.6a as a function of the composition. These values are in agreement with those derived from magnetisation data [28]. In fact, dHt /dTt is related to the strength of the magnetoelastic coupling: in these compounds, the value of ∆Siso measured when the transition is field-induced coincides with the value of ∆Siso measured when the transition is induced by the application of pressure [24]. Therefore, through the Clausius–Clapeyron equation (12.2), it is shown that dTt dPt ∆M =− . (12.4) ∆V dHt dTt According to (12.4), a strong magnetoelastic coupling yields a small value of dHt /dTt . We stress the linear dependence of dHt /dTt on x (see Fig. 12.6a), which is decreasing for the PM-to-FM transition (solid line in Fig. 12.6a), while it is increasing for the AFM-to-FM transition (dashed line in Fig. 12.6a). Both lines intercept at the composition range where the second-order transition disappears (0.2 < x < 0.24), in agreement with the phase diagram for these alloys [22]. The decrease of dHt /dTt with increasing x for the PM–FM transition indicates a strengthening of the magnetoelastic coupling. This may be explained by taking into account that the FM exchange interactions are stronger for increasing x, as suggested by the magnetic phase diagram [22, 25], where Tt increases linearly with x. Concerning the AFM–FM transition, the increase of dHt /dTt with x may be related to the fact that the transition occurs between two ordered states. Figure 12.6a thus summarises the behaviour of the magnetostructural transformation as a function of x, T and H. We have determined the field dependence of ∆S by proper integration of the calorimetric curves measured by sweeping T under fixed values of the magnetic field. Two examples which are representative of the obtained ∆S(H)

232

F. Casanova et al.

Fig. 12.6. (a) Slope of Ht (Tt ) calculated from DSC data under H. Solid line corresponds to the PM–FM transition, while dotted line corresponds to the AFM–FM transition. (b) Scaling of |∆S| at the first-order transition for the Gd5 (Six Ge1−x )4 alloys. Data corresponding to a large variety of applied fields and compositions are represented. Solid and open diamonds are from [26]. Symbols labelled with an H and H = 0 correspond, respectively, to measurements with/without magnetic field

curves are shown in the insets of Fig. 12.5 for samples with composition x = 0.1 and 0.3. For x ≤ 0.2, −∆S increases linearly with H, while for 0.24 ≤ x ≤ 0.5, −∆S decreases linearly with H. Interestingly, the absolute value of ∆S(H) curves corresponding to samples in the range of composition 0 ≤ x ≤ 0.5 scale in two branches (one increasing and the other decreasing) when they are plotted as a function of Tt , as shown in Fig. 12.6b. In this scaling there is only one parameter, Tt , which is tuned by both x and H. This demonstrates that this is the relevant quantity in determining the magnitude of |∆S| in

12 Entropy Change and Magnetocaloric Effect

233

these alloys, and proves that the magnetovolume effects due to H are of the same nature as the volume effects caused by substitution [27]. Each one of the quasi-linear branches of the scaling corresponds to one of the two ranges of composition in which AFM–FM and PM–FM magnetostructural transformations occur in the phase diagram. As a matter of fact, the shape of the scaling is governed through the Clausius–Clapeyron equation (12.2) by the dependence of dHt /dTt on x (see Fig. 12.6a) provided that ∆M decreases monotonically with Tt . Besides, the scaling is not a trivial consequence of the scaling of both ∆M and dHt /dTt , i.e., neither ∆M nor dHt /dTt scale with Tt , which gives further relevance to the scaling of |∆S| [27, 28]. Notice also that |∆S| extrapolates to zero at Tt = 0, as expected from the third law of the thermodynamics. In systems with large magnetoelastic coupling, as it is the case for Gd– Si–Ge alloys, the magnetostructural transformation can easily be induced by the applied magnetic field. Figure 12.7 shows examples for x = 0.1 of the calorimetric curves measured by increasing and decreasing the field at constant temperatures above Tt (H = 0). As for the mode in which temperature is swept, a proper integration of the signal corresponding to each peak yields the entropy change of the transition. In Fig. 12.7, results are compared for two sweeping rates which differ by one order of magnitude. Neither the shape of the calorimetric curves nor the calculated ∆S values differ significantly when both sets of measurements are compared. Figure 12.8a shows the evolution of the calorimetric curve when cycling repeatedly through the transition at 55 K with a sweeping rate of 1 kOe/min, in a sample with x = 0.05 in the virgin state at the start of the measuring

Fig. 12.7. Calorimetric curves recorded sweeping the magnetic field (increasing and decreasing H) in a Gd5 (Si0.1 Ge0.9 )4 sample at several fixed temperatures and for two different field rates

234

F. Casanova et al.

Fig. 12.8. (a) DSC curves recorded on increasing H at 1 kOe/min for different cycles in a Gd5 (Si0.05 Ge0.95 )4 sample. (b) Entropy change obtained from previous ˙ are given in the plots. data, as a function of the number of cycle. Field rates, H, Solid lines are a guide to the eye

process. Even at first glance, it is clear that the curve corresponding to the first cycle encloses a smaller area than the subsequent ones. It is worth noting that the fine and sharp peaks superimposed on the broad peak of the transition are not a consequence of experimental noise. Actually, they reveal the jerky character of the transition for alloys with x < 0.1, for which magnetostructural transformation takes place at relatively low temperatures at which the thermal activation does not blur the discontinuous character of the transition. This discontinuous behaviour is associated with avalanche-type dynamics, which originates from collective jumps between metastable states at mesoscopic scale [29]. Similar results have also been detected in Gd5 Ge4 from magnetisation measurements [30]. The superimposed peaks also evolve by cycling, reaching a pattern which is reproducible to some extent. This may be indicative of the athermal character of the transition [31]. That is to say, the transition is induced when the field reaches a critical value rather than being thermally activated. When a system is driven by an external parameter (the magnetic field in our case) through a first-order transition, the metastable-state path followed by the system depends on the disorder (microcracks, vacancies, dislocations). For an athermal transition the path can be reproduced from cycle to cycle provided that the disorder does not evolve [32]. Figure 12.8b shows the effect of cycling on |∆S| of the transition in the same sample. |∆S| increases strongly between first and fourth cycles and eventually gets saturated before the tenth cycle. This evolution is due to the formation of microcracks as a consequence of the continuous expansion/contraction of the crystallographic cell, until the system reaches a reproducible path. Besides, |∆S| values with increasing H are systematically larger than those obtained with decreasing H. This might be due to the hysteresis effects between increasing and decreasing branches.

12 Entropy Change and Magnetocaloric Effect

235

12.5 Conclusions The MCE associated with magnetostructural transformations is controlled by the interplay between magnetic and structural degrees of freedom and occurs simultaneously at multiple length scales. In the vicinity of the martensitic transition of Ni–Mn–Ga, it is observed from the magnetisation measurements that the MCE originates from (i) the magnetoelastic coupling on the mesoscopic scale between the magnetic moments and the martensitic variants, and (ii) the microscopic spin–phonon coupling which gives rise to the shift of the transition temperature with the applied field. The contribution from these different scales to the MCE changes with the intensity of the applied field and the composition of the alloy. In Gd–Si–Ge alloys, the giant MCE is controlled by the microscopic spin–phonon coupling (which is proportional to dHt /dTt ) since the transformation entropy change can be field-induced. ∆S can also be obtained directly from DSC operating under magnetic field. In these alloys, we observe that |∆S| scales with the transition temperature, proving that the magnetovolume effects due to H are of the same nature as the volume effects caused by substitution. The jerky character of the calorimetric curves for alloys with low transition temperatures, which indicates an avalanchetype dynamics, evolves with cycling until a reproducible pattern is reached, evidencing the athermal character of the transition. Interestingly, it has been found that the avalanche size distribution evolves with cycling from a subcritical distribution towards a critical (power-law) distribution [33]. |∆S| also varies with the first cycles through the transformation. Acknowledgements The financial support of the Spanish CICYT (MAT2003-01124 and MAT20013251) and Catalan DURSI (2001SGR00066) are recognised. F.C. and J.M. acknowledge DURSI for Ph.D. grants.

References 1. A.M. Tishin, Y.I. Spichkin: The Magnetocaloric Effect and its Applications (Institute of Physics Publishing, Bristol and Philadelphia, 2003) 2. V.K. Pecharsky, K.A. Gschneidner Jr.: J. Appl. Phys. 85, 5365 (1999) 3. K.A. Gschneidner Jr., V.K. Pecharsky: Annu. Rev. Mater. Sci. 30, 387 (2000) 4. A.M. Tishin, in: Handbook of Magnetic Materials, ed. by K.H.J. Buschow, Vol. 12 (North Holland, Amsterdam 1999) pp. 395–524 5. V.K. Pecharsky, K.A. Gschneidner Jr.: J. Magn. Magn. Mater. 200, 44 (1999) 6. F. Casanova, X. Batlle, A. Labarta, J. Marcos, L. Ma˜ nosa, A. Planes: Phys. Rev. B 66, 100 401(R) (2002) 7. V.K. Pecharsky, K.A. Gschneidner Jr.: Phys. Rev. Lett. 78, 4494 (1997)

236

F. Casanova et al.

8. H. Wada, Y. Tanabe: Appl. Phys. Lett. 79, 3302 (2001) 9. O. Tegus, E. Br¨ uck, K.H.J. Buschow, F.R. de Boer: Nature 415, 450 (2002) 10. V.K. Pecharsky, K.A. Gschneidner Jr.: Adv. Mater. 13, 683 (2001) and references therein 11. W. Choe, V.K. Pecharsky, A.O. Pecharsky, K.A. Gschneidner Jr., V.G. Young Jr., G.J. Miller: Phys. Rev. Lett. 84, 4617 (2000) 12. J. Marcos, L. Ma˜ nosa, A. Planes, F. Casanova, X. Batlle, A. Labarta: Phys. Rev. B 68, 094 401 (2003) 13. F.C. Lovey, V. Torra: Prog. Mater. Sci. 44, 189 (1999) 14. Q. Pan, R.D. James: J. Appl. Phys. 87, 4702 (2000) 15. H.D. Chopra, C. Ji, V.V. Kokorin: Phys. Rev. B 61, R14 913 (2000) 16. R.C. O’Handley: J. Appl. Phys. 83, 3263 (1998) 17. S.Y. Chu, A. Cramb, M. De Graef, D. Laughlin, M.E. McHenry: J. Appl. Phys. 87, 5777 (2000) 18. L. Ma˜ nosa, A. Planes, J. Zarestky, T. Lograsso, D.L. Schlagel, C. Stassis: Phys. Rev. B 64, 024 305 (2001) 19. J. Marcos, L. Ma˜ nosa, A. Planes, F. Casanova, X. Batlle, A. Labarta, B. Mart´ınez: Phys. Rev. B 66, 224 413 (2002) 20. F. Albertini, L. Morellon, P.A. Algarabel, M.R. Ibarra, L. Pareti, Z. Arnold, G. Calestani: J. Appl. Phys. 89(10), 5614 (2001) 21. L. Pareti, M. Solzi, F. Albertini, A. Paoluzi: Eur. Phys. J. B 32, 303 (2003) 22. V.K. Pecharsky, K.A. Gschneidner Jr.: Appl. Phys. Lett. 70, 3299 (1997) 23. J. Marcos, F. Casanova, X. Batlle, A. Labarta, A. Planes, L. Ma˜ nosa: Rev. Sci. Inst. 74, 4768 (2003) 24. L. Morellon, P.A. Algarabel, M.R. Ibarra, J. Blasco, B. Garc´ıa-Landa, Z. Arnold, F. Albertini: Phys. Rev. B 58, R14 721 (1998) 25. L. Morellon, J. Blasco, P.A. Algarabel, M.R. Ibarra: Phys. Rev. B 62, 1022 (2000) 26. A. Gigu`ere, M. F¨ olde` aki, B.R. Gopal, R. Chahine, T.K. Bose, A. Frydman, J.A. Barclay: Phys. Rev. Lett. 83, 2262 (1999) 27. F. Casanova, X. Batlle, A. Labarta, J. Marcos, L. Ma˜ nosa, A. Planes: Phys. Rev. B 66, 212 402 (2002) 28. F. Casanova, A. Labarta, X. Batlle, J. Marcos, L. Ma˜ nosa, A. Planes, S. de Brion: Phys. Rev. B 69, 104 416 (2004) 29. F.J. P´erez-Reche, M. Stipcich, E. Vives, L. Ma˜ nosa, A. Planes, M. Morin: Phys. Rev. B 69, 064 101 (2004) 30. V. Hardy, S. Majumdar, S.J. Crowe, M.R. Lees, D.M. Paul, L. Herv´e, A. Maignan, S. H´ebert, C. Martin, C. Yaicle, M. Hervieu, B. Raveau: Phys. Rev. B 69, 020 407(R) (2004) 31. F.J. P´erez-Reche, E. Vives, L. Ma˜ nosa, A. Planes: Phys. Rev. Lett. 87, 195 701 (2001) 32. J.P. Sethna, K. Dahmen, S. Kartha, J.A. Krumhansl, B.W. Roberts, J.D. Shore: Phys. Rev. Lett. 70, 3347 (1993) 33. F. Casanova, A. Labarta, X. Batlle, J. Marcos, L. Ma˜ nosa, A. Planes: Eur. Phys. J. B 40, 427 (2004)

13 Functional Magneto-Structural Materials: Summary and Perspectives P.-A. Lindg˚ ard

The idea to bring many disciplines in physics and materials science together in a fruitful way is a bold and important one. This is manifested in the present book Interplay of Magnetism and Structure in Functional Materials. The common theme is in a sense the real and unavoidable complexity of the structural problems found in seemingly quite unrelated fields. In my task to give an overview, my attempt will be to concentrate on broader aspects and on the perspectives – instead of dealing with rather smaller technical details of interest mainly for specialists among specialists. For the readers interested in further studies – beyond the chapters in this book – let me mention related topics briefly and attempt to give some relevant references as a starting point. It all began with the problem of understanding the martensitic phase transformation. When I was first introduced to that, now long time ago – coming from a field of statistical physics and critical phenomena – two things struck me (1) why was this considered as a special phase transition and (2) why did one not consider internal elastic strain modes, closely related to zone-boundary or other finite q-phonons, as variables in free energy expansions [1, 2]. Now, it is probably accepted by most that it is indeed a regular phase transition, displacive and diffusion less, usually weakly first order and with a lattice deformation as the order parameter. And it is the accommodation of that which is the problem because of the long-range character of the strain fields. This gives rise to the then usually much discussed precursor effects [3], which is not the main focus here. Neither is the remarkable nonmonotonous behavior of the hysteresis loop: the avalanche effect and self-organized critical phenomena, where analyses in terms of scaling laws and critical exponents come into play. Important progress has been made in this direction – in particular by the Barcelona group [4] – which I will not cover here. Part of this is covered in Chap. 2. In a paper by Hardy et al. [5] similar phenomena have been reported for manganites, of the type Pr1−x Cax Mn1−y My O3 (with x ∼ 0.5, y ∼ 0.05, and where M is a magnetic or nonmagnetic cation) in a field. A competition sets in between the magnetic energy promoting the ferromagnetic phase (FM) contrary to the antiferromagnetic (AFM) one, and the elastic energy

238

P.-A. Lindg˚ ard

associated with strains created at the FM/AFM boundaries. In addition an incubation time phenomenon is observed, such that a given jump or avalanche can occur after a certain waiting time under constant external conditions. Vives et al. [6] reported on a new effect found in the Cu–Al–Mn alloys: a disorder induced critical point at low temperatures as a result of a competition between ferro- and antiferromagnetic interactions. However, the main focus here is not on the phase transition aspects. The martensitic – or shape memory – transformation often occurs in quite complex alloys or nonstoichiometric materials. A physicist would rather study a model system to get a deep understanding of various problems – one at a time – if possible. In old times Zr served as such a model system [1, 2, 7] in which it was shown that the transition (from bcc to hcp) has to involve simultaneously both a uniform and an internal strain (corresponding to a transverse N -point [100] phonon). In other words the transition/reaction coordinate is (at least) two dimensional. It would come as no surprise that going from one point A in a mountainous landscape to another, B, can usually not be accomplished (without excessive efforts) following the straight line between A and B. One obviously would try to find a pass and be ready to make perhaps big detours – the bigger detour the smaller barrier. So there is a balance – and a transition path from A to B is not unique. One must even consider a distribution of possible paths. This fact is often overlooked in calculations of energy barriers in the martensitic problem, where we also deal with a rough “mountainous” energy landscape – or rather free-energy landscape. This resembles the situation found in nature’s own much more complex, and much more functional materials: the proteins [8]. Perhaps complexity and functionality are interrelated. However, there is still much work to do in understanding simple model systems – and a good candidate for a new system, AuZn, has been reported by Lashley et al. [9]. This has the advantage of being nonmagnetic – as Zr of course. It has a shape-memory transition from a cubic B2 (CsCl) to a trigonal P3 structure at ∼65 K, and may be ideal for studying the effects of nonstoichiometery on the martensitic transformation. It shows an essentially perfect second order transition with a λ-like specific heat, a transverse phonon softening at the N -point, etc. It can be obtained in perfect single crystals, which has allowed a de Haas–van Alphen determination of the Fermi surface [10]. This shows that there is a Fermi surface nesting feature matching the wave vector of the transition. That system seems to be nearly ideal – and understanding this in detail may be an important stepping stone toward understanding (also magnetic) shape memory alloys like NiTi [11] and Nix Al1−x [12], and the fashionable Ni2−x Mn1−x Ga [13], which all show large precursor effects at temperatures well above the martensitic transition temperature – as has been reported by Chernenko et al. [14], Pons et al. [15], and others. Shapiro et al. [16] performed numerous neutron scattering studies of systems in that family and other materials. And Ziebeck’s group performed neutron studies of the intricate Invar-problem [17] and of the Ni–Mn–Ga system [21]. Description

13 Functional Magneto-Structural Materials

239

of the neutron techniques and results obtained by these authors are reported in Chaps. 5 and 6, respectively. In the case of Zr, neutron scattering measurements of the phonon dispersion relations – and of soft modes – already gave very valuable insight into the transition mechanisms. Now, with the upcoming magneto-structural materials (promising for applications, because they can be influenced by a magnetic field), the neutrons will remain invaluable as a necessary experimental tool for gaining insight at the atomistic and the nanolevel. In addition, the possibilities are ripening to independently measure the spin and orbital components of the magnetic moment by X-ray/synchrotron methods, as discussed by Haskel et al. [18]. The magnetic shape memory systems come in two classes: those with uniaxial anisotropy, where the moment prefers a particular symmetry direction in the lattice; and those with planar anisotropy, which only confine the moments to certain preferred planes. For the shape memory effect to be useful it is important to control the development of the equivalent variants. For a transition from the bcc to hcp structure, there are six variants – and it can be more complicated as reported by Ahlers [19]. A selection can be made by a so-called “training process” of “smart” materials, during which deformation and transition are forced through a desired path, whereby certain variants can be promoted to nucleate earlier than others, in a complex process involving unknown features like defects, etc. in the structure. A much simpler control can be achieved by a magnetic field – in particular for the uniaxial materials. Kakeshita [20] reported on the uniaxial Fe∼3 Pd system, which transforms from a cubic to a tetragonal structure at 230 K, with one variant expanding along the applied field direction. Also studied is the planar Fe3 Pt system, which contracts in the applied field direction into a divariant tetragonal phase at 85 K. The transformation path for the rearrangement of the variants by the magnetic field and by an applied stress has been found to be the same (with an energy dissipated during the rearrangements of the variants of 260 kJ m−3 for Fe3.21 Pd). The effect depends strongly on the magneto-crystalline anisotropy parameter in the Hamiltonian term of the form −K Sz2 . For Fe∼3 Pd a value of K ∼ 350 kJ m−3 has been reported. It is highly important to theoretically understand the basic mechanisms behind the anisotropy, and to make accurate calculations. This, however, still represents a significant challenge. This has become very apparent in the seemingly totally unrelated system of use for magnetic cooling by exploiting the magneto-caloric effect. This relies on the entropy change, ∆S, at the phase transition. Increasing the magnetic order reduces the entropy, and the corresponding energy is released as heat in the lattice – and vice versa. This can be used to cool other attached objects. Pecharsky et al. [21] showed that ∆S can be significantly enhanced by magneto-structural effects. In 1997 a giant magneto-caloric effect was found [22] in the Gd–Si–Ge materials (more precisely R5 (Si4−x Gex ), where R is a rare earth element). This problem is discussed in Chap. 10. Although much is now understood – in particular that a shear between layers in the structure is important – a good deal of fundamental research remains to

240

P.-A. Lindg˚ ard

be done before a basic understanding is achieved. It is thus very difficult to understand why Gd5 (Si4−x Gex ) is one of the best materials – with as large an anisotropy constant as for Co. Gd, being an S-state ion, should on the contrary show only weak magneto-crystalline anisotropy. Here, clearly more theoretical foundation and quantitative calculations are needed. Perhaps a mechanism as that recently proposed for pure Gd applies here as well [23]. Labarta and collaborators [24] have reported on the influence that the composition x and the field H have on the phase transition and in particular on the entropy change ∆S. It has been concluded from the scaling behavior found for ∆S that the magneto-volume effects due to H are of the same nature as those due to atomic substitution, x. They further showed that the field induced maximum entropy change, ∆Smax , is even bigger per kg in the Ni–Mn–Ga systems (∼−30 J K−1 per kg at 1.5 T) than in the Gd5 (Si4−x Gex ) materials (∼−15 J K−1 per kg at 2.0 T) [25]. These results are reported in Chap. 11. When a material approaches the realm of applications, optimizations are called for – especially it seems necessary to reduce the magnetic field from the presently needed about 2 T (close to the maximum obtainable in permanent magnets) to lower values. An attempt to achieve that has been reported by Lewis et al. [26] who, by using thin films of Gd5 (Si4−x Gex ) coated by either Fe or Al, can reduce the needed field and even increase ∆S – equally well with both coatings. The basic reason remains obscure, but a good guess might be that the first-order transition starts by nucleation events in which the surface properties obviously play a large role – hence strain effects produced by the add atoms may create nucleation sites – irrespective of their magnetic properties. Clearly, there is room for lots of improvements along those lines. In a sense turning things upside down, Str¨ assle et al. [27] reported on how one in many systems one can obtain the magneto-caloric effect by applying a hydrostatic pressure. The basic mechanism can be much more diversified than the above discussed “anisotropy mechanism” and might include mechanisms ranging from pressure induced structural (Prx La1−x NiO3 ), magnetic (CeSb), valence [EuNi2 (Six Ge1−x )2 ] transitions, spin fluctuations (YMn2 ) to electron hybridization in Kondo systems (Ce3 Pd20 Ge6 ) – with potential effects as promising as in the above anisotropic magneto-caloric materials. Of course this cooling method is in principle the same as the conventional gas–liquid one. If one should make a choice between the various mechanisms, one tends to move into a space of – for physics – irrelevant variables like cost, efficiency, availability, noise, and other environmental effects, etc. Physics should be free to explore possibilities (not only avenues to prosperity, but also the impracticable paths) providing facts and inspiration for inventors. Unfortunately, the mentioned variables have a tendency to catch up with physics in the funding phase. The magneto-volume problem is of course related to the Invar-problem, which has been much studied by Wassermann et al. [28]. This issue is reviewed in Chap. 9. The name indicates “invariant”, namely an fcc Fe65 Ni35 alloy, which does not show thermal expansion (because of a counteracting

13 Functional Magneto-Structural Materials

241

volume magnetostriction) – an invention which brought the discoverer, C.E. Guillaume, the Nobel prize in 1920. However, in spite of almost 100 years of intensive research the basic mechanism has not been agreed upon – but Wassermann states that it most probably has to do with a charge transfer of electrons between orbitals of different symmetry (eg and t2g in ionic terms) and their nonbonding and antibonding characters; whereas a canting model of the moments cannot account for the phenomena [29]. Perhaps the meaning of “understanding” is relative: the more studies the stricter demands – now one might be able to solve some of the difficult (really metallic) problems with modern techniques and computers. This brings us to the electronic theory part. Clearly, in most cases one must consider the behavior of the electrons along with the structural and magnetic changes – on the same footing, a very difficult problem. Harmon [30] is one of the pioneers in the field, who first used the “frozen phonon” method to elucidate the martensitic transformation in for example Zr [1, 31]. Harmon et al. [32] reported on the advances, using the local density approximation, augmented by a special treatment of the 4f-electrons for the electron exchange and correlation, for the calculation of the properties of complex magnetic materials as Gd5 (Si2 Ge2 ) and Ni2 MnGa. The intriguing behavior of a Zoo of precursor phases (with different periodicity) might be understood (at least partly) in terms of a Fermi surface nesting vector which moves with the degree of magnetization. In NiTi systems the soft phonon modes can be found quite accurately and in accordance with experiments using molecular dynamics methods. An impressive work has been performed by Ayuela et al. [33] on the efforts to calculate a number of properties, like the field dependent shape memory effect, the phonon curves, the anisotropy energies, and the transition temperatures in the Ni–Mn–Ga systems, in which a field can induce as much as 10% strain. They use a quite different, atomistic approach than the above mentioned. A good agreement with the anisotropy constant has been obtained, just by taking into account the orbital character of the electrons and calculating the energy differences. A similar calculation should be done for the Gd5 (Si2 Ge2 ) system. It is not clear if changes in the Fermi surface with magnetization or if the entropy part is properly taken into account in this treatment. Apropos entropy, it is quite important that the various contributions are theoretically estimated, whether originating from the changes in the electronic entropy or the vibrational (and disorder) entropy or other sources. The again seemingly totally different subject of colossal magneto-resistance (CMR) materials of the manganite family, exemplified by Bi1−x Srx MnO3 , has been investigated by Frontera et al. [34]. Here an intimate interplay between magnetic and structural properties exists. However, in these basically insulating materials, the theory seems to be on better grounds with the Goodenough rules [35] than in the metallic magneto-structural alloys. These rules go a long way in explaining the interplay between the magnetic ordering and the charge ordering – although one additionally may have to consider a strong coupling

242

P.-A. Lindg˚ ard

between two neighboring Mn ions, which may then form a delocalized Zener polaron [36]. Many features of the effect of Bi substitution can be explained in terms of the Zener polaron model. In similar materials, Ibarra’s group [37] investigated the interplay between strong electron correlation and the lattice defects in mixed valence manganites. This is discussed in Chap. 3. They pointed out that an intrinsic inhomogeneous phase, even at the nanoscale, is found in which for example one would observe coexistence of ferromagnetic clusters in a paramagnetic insulating phase as well as coexistence of paramagnetic insulating regions and ferromagnetic metallic regions. A local disorder in these materials seems to be unavoidable. The balance is so delicate that one can observe pronounced isotope effects by exchanging 16 O by 18 O. This quite unexpected phenomenon has been shown by Egami [38], to occur both in the CMR ceramics and in the high temperature superconducting cuprates. It is a size not a phonon effect. He advocated for the importance of studying the ionic size effect on the CMR phenomena. To this a new – high-q – neutron scattering technique has been developed to very accurately determine the pair distribution function in solids. The general conclusion is that in these materials an intrinsic, unavoidable disorder occurs at the nanoscale level. If this – let us call it new – phase exists it requires rethinking of the thermodynamic treatment – namely an inclusion of the entropy corresponding to the intrinsic (static) disorder even at very low temperatures. This is an important challenge for theoreticians/statisticians in the years to come. This is covered in Chap. 4. As a theoretician, I am always impressed by the advances made in experimental techniques. De Graef et al. [39] nicely demonstrated the progress made in imaging magnetic domains using the Lorentz transmission electron microscope, showing for example domain walls and domain pinning on lattice defects in Ni2 MnGa. He also looked into the prospects for in situ imaging. The technique is explained in Chap. 7. Similar studies by Tsuchiya et al. [40] allowed an imaging of the tetragonal, intermediate, and cubic phases in the Ni–Mn–Ga alloys. Additional use of electron holography has been discussed by Murakami et al. [41], with applications to the Ni51 Fe22 Ga27 shape memory alloy and to the hole-doped manganites La1−x Srx MnO3 . A complex interplay between the magnetic and the structural domain structure has been observed. In the manganites it is found that both ferro- and antiferromagnetic phases coexist. Clearly, the real space experimental methods provide very valuable, complementary information to that obtained by the reciprocal space scattering techniques. To help putting some order into the complexity Shenoy et al. [42] formulated a classification of the disorder observed on various scales. He pointed out the importance of the different length scales and distinguishes among the macroscopic, the mesoscopic, the nanoscale, the atomic scale, and the electronic scale (as we saw one may need to go to even the nuclear scale – the O-isotope effect in manganites and cuprates). Using a Ginzburg– Landau approach he suggested a generic and elastic origin for the ubiquitous

13 Functional Magneto-Structural Materials

243

inhomogeneity phenomena: stripes, droplets, channels, twins domains, etc. The energy is lowered by the interweaving of competing low-temperature structural variants. This is developed in Chap. 1. That may in fact be why the martensitic transition so long took a special position among phase transitions. It is very important to have these scales in mind when discussing phenomena in as diversified a field as encompassed by the interplay of magnetism and structure. In a sense we seem to face the opposite of a Babylonian confusion, having the same language for different things. Cast´ an et al. [43] proposed a more detailed classification scheme for multistage displacive structural transformation in terms of the presence of, or absence of precursor modulated phases and applied it to the Ni–Mn–Ga, NiTi(Fe), Cu–Zn–Al and α-uranium systems. Classification schemes are good at creating new perspectives. The accommodation problem of the structural transitions has been dealt with also at an abstract mathematical level. Lexellent et al. [44] used a scale free, mathematical approach to discuss the joint effect of external field and stress with the Ni–Mn–Ga systems in mind. James et al. [45] reported (Chap. 8) on the interrelations between various strain tensors – and has solved the problems by referring to the properties of the SU(n) symmetry groups. It is inspiring to see the martensitic problem discussed in terms usually joggled within elementary particle physics. Perhaps, the above classifications should be even further extended. And perhaps our materials have rich enough physics that they may be used as experimental testing ground for the more exotic theories of the Universe. In conclusion, the chapters cover a wide area of materials science. They all bring something new or take a new angle on an old problem – and I have tried to amplify this principle, set out by the editors, by mentioning a couple of further, related investigations. Clearly the list of such works cannot be complete here. The feeling one is left with after the confrontation of the diverse phenomena is that physics is full of interesting phenomena and there is a lot to do, even at a very fundamental level. A particular flavor with all the materials covered here is further that apart from being intriguingly complex and of academic interest they are potentially useful for the society in a not too distant future. Acknowledgment I wish to thank the editors for bringing several of the discussed works to my attention, and also for useful comments.

References 1. Y.Y. Ye, Y. Chen, K.M. Ho, B.N. Harmon, P.A. Lindg˚ ard: Phys. Rev. Lett. 58, 1769–1772 (1987) 2. P.A. Lindg˚ ard, O.G. Mouritsen: Phys. Rev. Lett. 57, 2458–2461 (1986) 3. L.E. Tanner, A.R. Pelton, R. Gronsky: J. Phys. 43, 169–172 (1982)

244

P.-A. Lindg˚ ard

4. E. Vives, J. Ort´ın, Ll. Ma˜ nosa, I. R` afols, R. P´erez-Magran´e, A. Planes: Phys. Rev. Lett. 72, 1694–1697 (1994) 5. V. Hardy, A. Maignan, S. Hebert, C. Martin: Phys. Rev. B 67, 024401 (2003) 6. J. Marcos, E. Vives, Ll. Ma˜ nosa, M. Acet, E. Duman, M. Morin, V. Nov´ ak, A. Planes: Phys. Rev. B 67, 224406 (2003) 7. C. Stassis, J. Zarestky, N. Wakabayashi: Phys. Rev. Lett. 41, 1726–1729 (1978) 8. P.A. Lindg˚ ard: J. Phys.: Condens. Matter 15, s1779–s1786 (2003); H. Frauenfelder, B.H. McMahon: Annal. Phys. 9, 655–667 (2000) 9. J.C. Lashley, B.E. Lang, J. Boerio-Goates, B.F. Woodfield, T.W. Darling, A. Migliori, D. Thoma: J. Chem. Thermodyn. 34, 251–261 (2000) 10. R.D. McDonald, J. Singleton, P.A. Goddard, F. Drymiotis, N. Harrison, H. Harima, M.-T. Suzuki, A. Saxena, T. Darling, A. Migliori, J.L. Smith, J.C. Lashley: J. Phys. Condens. Matter 17, L69–L75 (2005) 11. N.B. Morgan, C.M. Friend: J. Phys. IV 11, 325–332 (2001) 12. R.D. Noebe, R.R. Bowman, M.V. Nathal: Int. Mat. Rev. 38(4), 193–232 (2001) 13. A. Sozinov, A.A. Likhachev, N. Lanska, K. Ullakko: Appl. Phys. Lett. 80, 1746–1748 (2002) 14. P. Mullner, V.A. Chernenko, G. Kostorz: Scripta Matter. 49, 129–133 (2003) 15. J. Pons, V.A. Chernenko, E. Cesari, V.A. L’vov: J. Phys. IV 112, 939–942 (2003) 16. A. Zheludev, S.M. Shapiro, P. Wochner, A. Schwartz, M. Wall, L.E. Tanner: J. Phys. IV 5, 1139–1144 (1995) 17. P.J. Brown, K.U. Neumann, K.R.A. Ziebeck: J. Phys.: Condens. Matter 13, 1563–1569 (2001) 18. D.R. Lee, S.K. Sinha, D. Haskel, Y. Choi, J.C. Lang, S.A. Stepanov, G. Srajer: Phys. Rev. B 68, 224409 (2003) 19. M. Ahlers: Philos. Mag. A – Phys. Condens. Matter Struct. Defects Mech. Prop. 82, 1093–1114 (2002) 20. T. Kakeshita, K. Ullakko: MRS Bull. 27, 105–109 (2002) 21. V.K. Pecharsky, A.P. Holm, K.A. Gschneidner Jr., R. Rink: Phys. Rev. Lett. 91, 197204 (2003) 22. V.K. Pecharsky, K.A. Gschneidner Jr.: Phys. Rev. Lett. 78, 4494–4497 (1997) 23. M. Colarieti-Tosti, S.I. Simak, R. Ahuja, L. Nordstrom, O. Eriksson, D. Aberg, S. Edvardsson, M.S.S. Brooks: Phys. Rev. Lett. 91, 157201 (2003) 24. F. Casanova, X. Batlle, A. Labarta, J. Marcos, Ll. Ma˜ nosa, A. Planes: Phys. Rev. B 66, 212402 (2002) 25. J. Marcos, Ll. Ma˜ nosa, A. Planes, F. Casanova, X. Batlle, A. Labarta: Phys. Rev. B 68, 094401 (2003) 26. L.H. Lewis, M.H. Yu, R.J. Gambino: Appl. Phys. Lett. 83, 515–517 (2003) 27. T. Str¨ assle, A. Furrer, Z. Hossain, C. Geibel: Phys. Rev. B 67, 054407 (2003) 28. P. Entel, H.C. Herper, E. Hoffmann, G. Nepecks, E.F. Wassermann, M. Acet, V. Crisan, H. Akai: Philos. Mag. B – Phys. Condens. Matter Stat. Mech. Electron. Opt. Magn. Prop. 80, 141–153 (2000) 29. M. van Schilfgaarde, I.A. Abrikosov, B. Johansson: Nature 400, 46–49 (1999) 30. B.N. Harmon, W. Weber, D.R. Hamann: J. Phys. 42, 628–630 (1981) 31. Y. Chen, C.L. Fu, K.M. Ho, B.N. Harmon: Phys. Rev. B 31, 6775–6778 (1985) 32. B.N. Harmon, V.N. Antonov: J. Appl. Phys. 93, 4678–4685 (2003); Y. Lee, J.Y. Rhee, B.N. Harmon: Phys. Rev. B 66, 054424 (2002) 33. A.T. Zayak, P. Entel, J. Enkovaara, A. Ayuela, R.M. Nieminen: Phys. Rev. B 68, 132402 (2003)

13 Functional Magneto-Structural Materials

245

34. C. Frontera, J.L. Garc´ıa-Mu˜ noz, M.A.G. Aranda, M. Hervieu, C. Ritter, Ll. Ma˜ nosa, X.G. Capdevila, A. Calleja: Phys. Rev. B 68, 134408 (2003) 35. J.S. Zhou, J.B. Goodenough: Phys. Rev. B 68, 054403 (2003) 36. C. Zener: Phys. Rev. 82, 403–405 (1951) 37. D. Rybicki, C. Kapusta, P.C. Ridi, C.J. Oates, M. Sikora, D. Zajac, J.M. De Teresa, C. Marquina, M.R. Ibarra: Acta Phys. Pol. A 105, 183–188 (2004) 38. T. Egami: Physica C 364, 441–445 (2001) 39. M. De Graef, Y. Kishi, Y. Zhu, M. Wuttig: J. Phys. IV 112(2), 993–996 (2003) 40. K. Tsuchiya, A. Tsutsumi, H. Nakayama, S. Ishida, H. Ohtsuka, M. Umemoto: J. Phys. IV 112(2), 907–910 (2003) 41. Y. Murakami, D. Shindo, K. Oikawa, R. Kainuma, K. Ishida: Appl. Phys. Lett. 82, 3695–3697 (2003) 42. A.R. Bishop, T. Lookman, A. Saxena, S.R. Shenoy: Eur. Phys. Lett. 63, 289–295 (2003); S.R. Shenoy, T. Lookman, A. Saxena, A.R. Bishop: Proceedings of the NATO Advanced Research Workshop (2004) and 30th International School of Solid State Physics, Erice, Sicily, 4–10 October 2003, in Symmetry and heterogeneity in high-temperature superconductors, ed. by A. Bianconi (Kluwer, Amsterdam, 2003) 43. T. Cast´ an, A. Planes, A. Saxena: Phys. Rev. B 67, 134113 (2003) 44. L. Hirsinger, C. Lexcellent: J. Phys. IV 112(2), 977–980 (2003) 45. A. DeSimone, R.D. James: J. Phys. IV 112(2), 969–972 (2003)

Index

Acoustic emission, 40 Adiabatic demagnetisation, 223 Advanced materials magnetocaloric, 199 Aharonov–Bohm relation, 143 Amplitude image, 147 Anti-Invar, 177, 181, 183, 187 Antiferromagnet, 3, 19 Antiferromagnetic Invar-effect, 178 Aperture function, 145 Astigmatism, 145 Athermal, 28, 37, 39–41, 43–45 Attenuation function, 145 Austenite/martensite interface, 167, 168 Austenitic, 141, 148, 152, 156 Avalanche(s), 39–45 Beam coherence, 145 divergence, 145, 146 Bend contour, 149, 153–155 Bragg angle, 142 Calorimetric curves, 230, 231, 233, 235 measurements, 223, 229 Clausius–Clapeyron, 224, 231, 233 CMR, 52, 59, 61–63 Co–Ni–Al, 33, 35, 45 Co–Ni–Ga, 33, 45 Cofactor conditions, 168–171, 174 Colosal magnetoresistivity (CMR), 75–80, 83, 84, 89 Compatibility, 28, 37, 38, 44, 45

Complex functional materials, 1 Compositional disorder, 15, 21 Constitutive response, 2 Convergent wall image, 143 Convolution operator, 145 Cooling power, 200 Coupling magnetoelastic, 224–228, 230, 231, 233, 235 mesoscopic, 224, 225, 228, 235 microscopic, 226, 228, 235 spin–phonon, 224, 226, 228, 235 Critical subcritical, 42 supercritical, 42 Cross-tie wall, 150 Crystal structure, 203, 206, 207, 209, 211, 213–217 Crystallography, 204, 206, 215, 216 Cu–Al–Ni, 30, 41, 42, 45 Cuprates, 3–6, 15–17, 19 Curie temperature, 16, 153, 154 CuZnAl, 163 Cu–Zn–Al, 31, 32, 39–41, 45 Damping function, 145 Deflection angle, 142, 143, 145 Defocus, 143, 145–147 Demagnetization field, 149, 150 Density of States, 183, 193 Deviatoric, 8–11, 15, 18, 20 Differential phase contrast, 143

248

Index

Dislocation, 141, 150 Disorder, 27, 28, 31, 36–38, 40, 43–45 Dispersion curves, 94–97, 99–104, 107–110 Distortion matrix, 161, 163–172 Divergent wall image, 143 Domain wall, 141, 143, 146, 149–156 Driving field, 37, 40, 43 rate, 39–42, 45 Elastic anisotropy, 30–32 constants, 29, 30, 34, 35 Elastic anisotropy, 7 compatibility, 6 constants, 180, 182 Electron diffraction pattern, 153, 154 microscopy, 141 wavelength, 142 Electron–phonon coupling, 84 Electronic theory anisotropy, 241 colossal magneto resistance (CMR), 241 Fermi surface, 241 frozen phonon, 241 Electrostatic potential, 143 Elinvar, 177, 179, 180 Energy barriers, 37, 38, 44 Entropy change, 201, 212, 217–219, 223–231, 233–235 Ericksen–Pitteri neighborhood, 165, 170 Fe3 Pd(Pt) Gd–Si–Ge, R5 (Si4−x Gex ), 239 FePd, 4, 8, 9, 21 FeRh, 202, 203 Ferroelastic, 4–6, 8, 9, 12–16 domain, 153 Ferroic, 5 Ferromagnet, 5 Ferromagnetic, 225 ordering, 199, 203, 204, 211, 216, 217, 219 shape memory alloy(s), 148, 156, 212 Invar-effect, 178

Fluctuations, 28, 31, 32, 38, 40, 41 Flux quantum, 144 Fourier space, 144 Fourier transform operator, 145 Free energy, 4, 5, 7–9, 11–14, 19–22 Fresnel mode, 143 Gd, 200, 201, 203, 212 Gd5 (Si2 Ge2 ), 202–205, 213, 215–219 Gd5 (Si4−x Gex ), 199, 202–206, 210, 213–215, 217 Gd5 Si4 , 203–205, 209, 211, 213–215 Geometric nonlinearity, 8, 9 Ginzburg–Landau, 8 theory, 2 Habit plane, 15 Heusler alloys, 161 (Hf0.83 Ta0.17 )Fe2±x , 202 HS–LS transition, 192, 195 Hyperfine field, 187, 189, 190 Hysteresis, 37–39, 43, 45, 159, 161–164, 168, 171–174, 225, 234 Imaging energy filter, 143 Inclusion, 12, 13 Inelastic neutron scattering, 93, 96, 99, 110 Instability, 29, 30, 34 Interaction constant, 144 Intermetallic compounds, 49, 55 Internal strain Fermi surface, 238 Invar-problem, 238 reaction coordinate, 238 uniform strain, 238 Intrinsic heterogeneity, 4 Invar, 177–185, 187–196 Ionic size effect, 76, 80 Irreversibility, 203, 210 Isotope effects accommodation problem, 243 classification scheme, 243 electron holography, 242 electron microscope, 242 length scales, 242 real space experimental methods, 242 reciprocal space scattering techniques, 242

Index Jahn–Teller, 5, 7 Jahn–Teller effect, 76 (La1−x Cax )MnO3 , 202, 203 La1−x Srx MnO3 , 242 LaCaMnO, 16 Laplacian, 146, 148, 153, 155 LaSrCuO, 16 Linear imaging model, 145 Local structure, 78, 82 Lorentz force, 142, 156 microscopy, 141, 143, 145 transfer function, 145 LTEM, 141, 148, 156 Magnetic cooling, 199, 200 domain(s), 224, 225, 227 field, 33, 34, 39, 223–231, 233–235 field induced temperature variation(s), 199 materials, 28, 40 refrigeration, 223 refrigerator, 199, 200, 203 transition, 223, 224, 230 Magnetic domain, 141, 143, 146, 148, 150, 152, 154–156 wall, 1 induction, 142, 149, 150 tweed, 154 vector potential, 143 Magnetisation, 223–229, 231, 234, 235 vector, 148 Magnetism, 100, 109 Magnetocaloric compounds, 200 effect, 199–202, 205, 206, 212, 213, 215–219, 223, 229 giant, 199, 205, 206, 212, 215, 217, 219 Magnetocrystalline anisotropy, 148, 152, 156 Magnetoelastic, 4, 18, 34, 36, 37, 45 material, 141 tweed, 141, 154–156 Magnetoresistance, 3, 15, 16, 49, 50, 52, 53, 55–57, 59, 61, 62, 68, 70 Magnetostatic energy, 150

249

Magnetostriction, 49, 52, 54–58, 61–64, 68–71 Magnetostructural, 28, 42 transformation, 224, 225, 230, 231, 233–235 transition, 199, 205, 212, 213, 215, 219 Manganese persovskites, See Manganites Manganites, 3, 5, 6, 15–17, 19 Martensite, 4, 13–15, 17, 109, 110 Martensitic, 148, 152–156 materials, 29, 30 transition, 28, 29, 33, 224, 225, 235 variant(s), 224–226, 228, 235 Metastable states, 43 Microscope transfer function, 145 Microstrain, 7 Modulated phase, 34, 35 structure, 29, 34 Moment–volume fluctuations, 180 Multiferroic, 3–7, 16–18, 21, 22 materials, 160 Multifunctional material(s), 3 Multiscale, 2, 225 Neutron Scattering, 116, 118, 119, 121, 135, 136 anisotropy, 239 entropy change, 239 magneto-caloric, 239 magneto-structural, 239, 241 variant, 239 X-ray/synchrotron scattering, 239 Ni2 MnGa, 212 Ni51 Fe22 Ga27 , 242 Ni–Al, 30–32, 45 Ni–Mn–Ga, 33, 35, 42, 43, 45, 224, 225, 229, 235 NiTiCu, 163, 171, 173 NiTiCuPd, 171, 172 Order parameter, 4–6, 11, 16, 17, 22 Out-of-focus mode, 142, 143, 152–155 Pair-density function (PDF), 76–81, 250 Paramagnetic, 230 neutron scattering, 194 Percolative metallicity, 17

250

Index

Phase contrast function, 145 reconstruction, 141, 147, 148, 156 shift, 141, 143, 145, 156 transition, 113, 116, 117, 120, 128, 129, 134–138 accommodation problem, 237 Extended, 27, 37, 38 First-order, 27, 37, 38 Ideal, 27 internal strain, 237 transformations, 49, 50, 52, 53, 55 uniform strain, 238 Phonon(s), 100–103, 105, 109 branch(es), 29–31, 33 Plane wave, 144 Polaron, 19, 20, 76, 79, 80 Polymorphs, 209, 214 Power law exponent, 42 Pr1−x Cax Mn1−y My O3 , 237 Precursor(s), 28, 29, 31, 33, 35 Premonitory, 29, 45 Pressure Invar-problem, 240 Kondo system, 240 magneto-volume, 240 Pretransitional, 27–29, 33, 37, 45 Prx La1−x NiO3 Ce3 Pd2 0Ge6 , 240 CeSb, 240 EuNi2 (Six Ge1−x )2 , 240 Fe65 Ni35 , 240 YMn2 , 240 Pseudogap, 16 R5 T4 , 203, 205, 212, 217, 219 Random field, 43 Relaxor ferroelectric, 3, 4, 15 Renormalization group, 1, 2 Reversibility, 159, 162, 164, 168, 169, 171, 174, 210 SANS, 60, 61, 63–65, 70 Scaling, 232 Shape change, 212 memory, 93, 95, 102–104, 110, 111, 113, 114, 128, 129, 137, 138

alloy(s), 3, 4, 29, 33, 44, 45, 212, 224 effect, 33, 225, 228 Shear displacement, 209 Sm5 Ge4 , 204–206, 209, 211 Softening, 29–33, 35, 36, 45 Soliton, 5, 14 Special lattice parameters, 159–163, 169, 172, 174 Specific heat, 180–183, 191 Spherical aberration, 145 Spin–charge stripes, 76, 83, 84 Strain, 152, 154, 156 tensor, 4, 6–9, 21 Stripes, 3, 16 Structural change, 199, 205, 206, 209, 210, 212, 215, 217, 219 modulations, 33 transition, 27, 36 TbMnO, 17 Temperature change, 201 span, 200 Temperature change, 223 Texture(s), 29 Thermal expansion linear, 206 volumetric, 206 Thermoelectric material, 174 Thermomagnetic material, 174 Transformation, 152–154 first-order, 205, 215, 216, 219 martensitic-like, 205, 209, 214, 215 polymorphic, 219 second-order, 215, 216 structural, 205, 207, 211, 213, 214, 217, 219 Transition magnetostructural, 199, 205, 212, 213, 215, 219 first-order, 224, 232, 234 second-order, 230, 231 Transmission electron microscopy, 141 Transport-of-intensity equation, 141 Tweed, 31, 32, 35–37, 45 Tweed, 4, 5, 13, 15–17, 153–155

Index Twins(s), 225, 227, 228 boundary, 155 system, 167–171

Unit cell dimensions, 206, 207, 210–212 “Unlikely” materials, 160

Vibronic mechanism, 87, 89

X-ray diffraction, 211, 213 XMCD, 190, 191 Zr AuZn, 238 Bi1−x Srx MnO3 , 241 Gd5 (Si2 Ge2 ), 241 Nix Al1−x , 238 Ni2 MnGa, 241 Ni2−x Mn1−x Ga, 238 NiTi, 238

251